Camp HD Thesis

The Pennsylvania State University
The Graduate School

Department of Acoustics
AN OPTIMAL GEAR DESIGN METHOD FOR MINIMIZATION

OF TRANSMISSION ERROR AND VIBRATION EXCITATION
A Dissertation in
Acoustics
by
Cameron P. Reagor
c 2010 Cameron P. Reagor

°
Submitted in Partial Fulfillment

of the Requirements
for the Degree of
Doctor of Philosophy
May 2010
The dissertation of Cameron P. Reagor was reviewed and approved* by the following:
William D. Mark
Senior Scientist, Applied Research Laboratory
Professor Emeritus of Acoustics
Dissertation Adviser
Chair of Committee
Stephen A. Hambric
Senior Scientist, Applied Research Laboratory
Professor of Acoustics
Gary H. Koopmann
Distinguished Professor of Mechanical Engineering
Victor W. Sparrow
Professor of Acoustics
Interim Chair of the Graduate Program in Acoustics
*Signatures are on file in the Graduate School.

iii
Abstract
Fluctuation in static transmission error is the accepted principal cause of vibration

excitation in meshing gear pairs and consequently gear noise. More accurately, there are
two principal sources of vibration excitation in meshing gear pairs: transmission error
fluctuation and fluctuation in the load transmitted by the gear mesh. This dissertation
formulates the gear mesh vibration excitation problem in such a way that explicitly
accounts for the aggregate contributions of these excitation components. The Fourier
Null Matching Technique does this by imposing a constant value on the transmission error
and solving for the requisite contact region on a tooth surface that yields a constant load
transmitted by the gear mesh. An example helical gear is created to demonstrate this
approach and the resultant compensatory geometry. The final gear tooth geometry is
controlled in such a way that modifications to the nominal involute tooth form exactly
account for deformation under load across a range of loadings. Effectively, the procedure
adds material to the tooth face to control the contact area thereby negating the effects of
deformation and deviation from involute. To assess the applicability of the technique, six
deformation steps that correlate to loads ranging from light loading to the approximate
full loading for steel gears are used. A nearly complete reduction in transmission error
fluctuations for any given constant gear loading should result from the procedure solution.
This overall method should provide a substantial reduction in the resultant vibration
excitation and consequently, noise.
iv
Table of Contents
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Chapter 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2. Previous Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 History of Gearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Noise in Gearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Current Gear Noise State of the Art . . . . . . . . . . . . . . . . . . 8
2.4 Development of a Complete Solution . . . . . . . . . . . . . . . . . . 8
2.5 Present Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Chapter 3. Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1 Gearing Constructs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Mathematical Preliminaries . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.1 The Involute Curve . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.2 Angle Relations . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.3 Gear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.4 Regular Gear Relations . . . . . . . . . . . . . . . . . . . . . 24
3.3 Transmission Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4 The Tooth Force Equation . . . . . . . . . . . . . . . . . . . . . . . . 26
Chapter 4. The Main Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.1 Bounding the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2 The Role of Convolution in Gear Noise . . . . . . . . . . . . . . . . . 32
4.3 Convolution’s Relation to Gear Geometry . . . . . . . . . . . . . . . 36
4.4 Developing the Procedure . . . . . . . . . . . . . . . . . . . . . . . . 37
4.5 Applying Fourier Principles . . . . . . . . . . . . . . . . . . . . . . . 39
4.6 Extending the Contact Region . . . . . . . . . . . . . . . . . . . . . . 41
4.6.1 The First Step . . . . . . . . . . . . . . . . . . . . . . . . . . 41
v
4.6.2 Continuing the Expansion . . . . . . . . . . . . . . . . . . . . 42

4.7 Fourier Null Matching Technique . . . . . . . . . . . . . . . . . . . . 44
4.8 Practical Work Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Chapter 5. The Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.1 Test Gear Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 The Test Gear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Compliance and Loading . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.4 Contact Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.5 Endpoint Modification . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.6 Discussion of Computed Contact Regions . . . . . . . . . . . . . . . 89
Chapter 6. Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 92

6.1 The Fourier Null Matching Technique . . . . . . . . . . . . . . . . . 92
6.2 The Practicality of the Tooth Modification . . . . . . . . . . . . . . . 93
6.3 Limitations of the Analysis . . . . . . . . . . . . . . . . . . . . . . . 93
6.4 Extending the Process . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.5 Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Appendix A. Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Appendix B. Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

B.1 Simplifying the Load Angle Equation . . . . . . . . . . . . . . . . . . 100
B.2 Radius of Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Appendix C. The Analysis Primary Code . . . . . . . . . . . . . . . . . . . . . . 104

C.1 Main . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
C.2 Gear Design Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
C.3 Parameter Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
C.4 Initiate the Finite Element Programs . . . . . . . . . . . . . . . . . . 110
C.5 Initialize the Finite Element Analysis . . . . . . . . . . . . . . . . . . 110
C.6 Set the Location Constants . . . . . . . . . . . . . . . . . . . . . . . 111
C.7 Open Saved Progress File . . . . . . . . . . . . . . . . . . . . . . . . 113
C.8 Graph the Legendre Loading Coefficients . . . . . . . . . . . . . . . . 115
C.9 Develop the Element Positions . . . . . . . . . . . . . . . . . . . . . 116
C.10 Map the Loading Profile on the Current Element Set . . . . . . . . . 119
C.11 Find the Hertzian Width . . . . . . . . . . . . . . . . . . . . . . . . . 121
vi
C.12 Mesh the Finite Element Model . . . . . . . . . . . . . . . . . . . . . 123

C.13 Perform the Finite Element Analysis . . . . . . . . . . . . . . . . . . 139
C.14 Apply Forces to the Finite Element Model . . . . . . . . . . . . . . . 140
C.15 Execute the Finite Element Analysis . . . . . . . . . . . . . . . . . . 142
C.16 Read the Finite Element Data . . . . . . . . . . . . . . . . . . . . . . 143
C.17 Prepare the Finite Element Data for Plotting . . . . . . . . . . . . . 146
C.18 Solve the Compliance Matrix . . . . . . . . . . . . . . . . . . . . . . 147
C.19 Reset the Loading Data . . . . . . . . . . . . . . . . . . . . . . . . . 149
C.20 Save the Critical Iteration Data . . . . . . . . . . . . . . . . . . . . . 150
C.21 Determine the Boundary of the Unmodified Region . . . . . . . . . . 150
C.22 Determine the Edge Extension . . . . . . . . . . . . . . . . . . . . . 153
C.23 Determine Step Size . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
C.24 Apply Default Loading . . . . . . . . . . . . . . . . . . . . . . . . . . 162
C.25 Open A Saved Data Set . . . . . . . . . . . . . . . . . . . . . . . . . 163
C.26 Rotate the Node Plane . . . . . . . . . . . . . . . . . . . . . . . . . . 164
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
vii
List of Tables
5.1 Test Design Gear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.2 Test Design Gear Continued . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.3 Summary of αΩ(N/µm). . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4 Summary of Line of Contact End Points. All values in m. . . . . . . . . 77
A.1 Summary of General Gearing Notation . . . . . . . . . . . . . . . . . . . 97

A.2 Summary of Transmission Error Notation . . . . . . . . . . . . . . . . . 99
A.3 Summary of Optimization Notation . . . . . . . . . . . . . . . . . . . . . 99
viii
List of Figures
2.1 An example of meshing gear teeth is shown. The global pressure angle,
φ, the base radius, Rb , and the pitch radius, R are indicated. . . . . . . 4
2.2 An example transmission error profile is shown. The horizontal coordi-
nate can be thought of as roll distance as the gear rotates. The vertical
coordinate is the deviation from perfect transfer of rotational position
expressed at the tooth surface. 1.2 cycles of gear rotation are shown. . . 5
2.3 An example Fourier spectrum of the transmission error is shown. The
horizontal coordinate is the gear rotational harmonic. The vertical co-
ordinate is the amplitude of each rotational harmonic. The gear has 59
teeth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Crowning: Typically a single lead modification and a single profile mod-
ification, (a), linearly superimpose to give a crowned deviation from per-
fectly involute, (b). Material, as described in this deviation from involute,
is removed from the perfect involute tooth, (c), in a direction normal to
the tooth surface thereby forming the crowned tooth illustrated in (d). . 7
2.5 From left to right: Given a finite element mesh of a gear tooth, any
element in that mesh that has an applied load also has size constraints.
Any applied lineal load along the length of the element creates a Hertzian
contact region b and has a unique ratio of element height c to width e
associated with it that is required to accurately predict b. . . . . . . . . 10
3.1 Standard gear diagram showing the analysis gear below and the mating
gear above it. The base plane projection of this helical gear pair is shown
above the mating gears. From Mark [5]. . . . . . . . . . . . . . . . . . . 12
3.2 The involute curve, I, is traced by the end of a string unwrapped from
a cylinder (with radius Rb ). The roll angle, ², is the angle swept by
the point of tangency, T , for the string with the cylinder as the string
unwraps. The pressure angle, φ, is the angle of force for meshing gears. φ
is measured against the pitch plane (the horizontal plane perpendicular
to the plane containing the gear axis). The construction pressure angle,
φi , is equal to φ when the “string” is unwrapped to the pitch point, P 0 . 14
ix
3.3 Dual Construction: for a set of local pressure angles, φi , locations on a

single tooth profile or multiple profiles rotating through space can satisfy
the resultant geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 The four gear planes: the transverse plane in blue, the pitch plane in
green, the axial plane in orange and the base plane in red. The mating
gear would be located directly above the gear cylinder for the gear of
analysis shown in dark blue. . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 A gear tooth (red) with its equivalent rack shown in blue and line of
contact in black. The green tooth profile intersects the line of contact at
the pitch point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6 An involute curve (black) is commonly illustrated originating from the
horizontal or x-axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.7 θ0 , as measured on the face of the equivalent basic rack tooth, is the angle
between the line of contact (magenta) and the base of the rack tooth.
The pressure angle, φ, is measured in the transverse plane. As measured
in the pitch plane, the pitch cylinder helix angle, ψ, is measured between
the tooth base normal and the transverse plane. φn is the elevation of
the tooth face normal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.8 While the roll angle, ², can be broken up into its components, φi and
θi , the pressure angle, φi , can be further divided at the tooth thickness
median into βi and ιi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.9 The force that is incident on a gear tooth can be broken into three com-
ponents. The helix and pressure angles are evident. . . . . . . . . . . . 21
3.10 Gear tooth ranges, L and D, relative to tooth rotation. . . . . . . . . . 23
4.1 Frequency domain theoretical gear noise is bound to integer multiples of

the primary gear rotational harmonic (n = 1). A vast majority of the
noise power in real world gears is found here as well. . . . . . . . . . . 32
4.2 A unit square function (a) has a frequency spectrum (b) with zeros at
integer harmonics. A unit triangle function (c) which is the convolution
of two square functions also has a frequency spectrum (d) with zeros at
the same integer harmonics. Note the difference in the falloff of the two
spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 A narrow square function is convolved with a triangle function to create
a new rounded triangular function. . . . . . . . . . . . . . . . . . . . . 34
x
4.4 The convolution of two square functions can produce a triangle function
or a trapezoidal function (a). Either is a “second order” function that
has a Fourier transform with asymptotic falloff of f −2 in frequency (b).
A “third order” convolved function (c) will produce a Fourier asymptotic
falloff on the order of f −3 (d). . . . . . . . . . . . . . . . . . . . . . . . 35
4.5 A square function of width ² convolves with a triangular function of width
2∆ to form a new function with a third order Fourier asymptotic falloff. 36
4.6 The black box denotes the available area on a tooth surface. The grey
box denotes the nominal Qa = Qt = 1 contact region. This region is only
valid for very light loading. As the line of contact travels from s = −∆ to
s = ∆, the practical width of line within the grey-boxed contact region
behaves like a triangular function. The tooth root is located at small ²
and the tooth tip is located at the upper bound of ². . . . . . . . . . . 40
4.7 The unmodified region of a line of contact, denoted by the flat area
at loading u00 , are extended on either side by some distance, θ1 . This
distance corresponds to the first design loading, u01 . . . . . . . . . . . . . 42
4.8 The line of contact includes two extensions beyond the unmodified con-
tact region. The first step extends the range by θ1 on either end and the
second step extends the previous line length by θ2 at each end. . . . . . 43
4.9 An example Fourier transmission error spectrum is shown through the
25th tooth harmonic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.10 An example continuous frequency noise spectrum is illustrated where
the noise energy is centered on integer multiples of the tooth meshing
fundamental. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.11 The frequency domain envelope of a Fourier Null Matching Technique
load profile is shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.12 An illustration of a Fourier Null Matching Technique resultant transmis-
sion error spectrum is shown. . . . . . . . . . . . . . . . . . . . . . . . . 47
4.13 An illustration of several lines of contact. The outer box represents the
available area on a tooth face. The inner box represents the Qa = Qt = 1
area which bounds the line of contact for −∆ ≤ s ≤ ∆. The step size
between lines of contact is ∆/4 in s while the horizontal coordinate is y
and the vertical coordinate is z. . . . . . . . . . . . . . . . . . . . . . . . 48
xi
4.14 The lower triangular figure represents the force transmitted by a single
tooth pair as they come in and out of contact. More completely, the total
force per unit depth deformation over a line of contact as a function of
s for each tooth-to-tooth cycle is represented as this triangle over the
range −∆ ≤ s ≤ ∆. The upper figure is the superimposed transmitted
forces of all tooth pairs. The total gear mesh transmitted load (sum of
the superimposed triangles) remains constant as the gear rotates. . . . . 49
4.15 A conceptual illustration of the nominal “lightly loaded” contact region
(bounded in blue) that satisfies Equation 4.13. Note that the initial rect-
angular region is not the bounds of the zone of contact that corresponds
to a triangular transmitted force profile. Several lines of contact are
shown in red. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.16 A conceptual illustration of the first expanded contact region (narrow
blue). The contact region that corresponds to regular transmitted force
profile (e.g. a triangle) is not regularly shaped itself (e.g. not rectangular). 51
5.1 A rendered example of the test gear. . . . . . . . . . . . . . . . . . . . 54

5.2 A closer view of the test gear showing the pitch point, P , and point of
tangency, T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3 A standard isometric view of the test gear. The pitch point, P , and point
of tangency, T are indicated. . . . . . . . . . . . . . . . . . . . . . . . . 55
5.4 An example of a finite element model used for the compliance analysis. 56
5.5 A close up of a finite element model used for the compliance analysis.
This model is for s = 0.8∆ . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.6 The loading curves for s = −0.008085 m . . . . . . . . . . . . . . . . . . 58
5.7 The loading curves for s = −0.0071866 m . . . . . . . . . . . . . . . . . 58
5.9 The loading curves for s = −0.00539 m . . . . . . . . . . . . . . . . . . . 59
5.12 The loading curves for s = −0.002695 m . . . . . . . . . . . . . . . . . . 61
5.15 The loading curves for s = 0 m . . . . . . . . . . . . . . . . . . . . . . . 62
5.16 The loading curves for s = 0.00089833 m . . . . . . . . . . . . . . . . . . 63
xii

5.18 The loading curves for s = 0.002695 m . . . . . . . . . . . . . . . . . . . 64
5.21 The loading curves for s = 0.00539 m . . . . . . . . . . . . . . . . . . . . 65
5.24 The loading curves for s = 0.008085 m . . . . . . . . . . . . . . . . . . . 67
5.25 The nominal loading case results in a perfect triangle. The remain-
ing load steps deviate slightly from a triangular profile. Note that the
rounded triangle is not immediately apparent due to resolution in s. . . 69
5.26 The bounds of each line of contact and its position on the tooth face is
shown in the same orientation as Figure 4.6 for deformation step u =
1µm. Blue box denotes Qa = Qt = 1. The lines of contact vary from
s = −0.9∆ on the lower right to s = 0.9∆ on the upper left with a step
between of ∆s = ∆/10. The tip of the tooth is positive z and the root
is negative z. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.27 The bounds of each line of contact for deformation step u = 5µm. Blue
box denotes Qa = Qt = 1. The lines of contact vary from s = −0.9∆
on the lower right to s = 0.9∆ on the upper left with a step between of
∆s = ∆/10. The tip of the tooth is positive z and the root is negative z. 71
xiii
5.32 The estimated contact region for the final geometry is seen. Note the well-
behaved portion of the tooth: negative s (lower right) for deformation
steps from 5 µm to 25 µm. . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.33 Each line of contact is constrained to have symmetric adjustments to the
length of the line of contact for each deformation step. The positive-y
extension profile is a mirror of the negative-y deformation profile. . . . . 78
5.34 The positive y end point modifications for s = −0.00808 m for all defor-
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.42 The positive y end point modifications for s = −0.000898 m for all de-
formation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.43 The positive y end point modifications for s = 0 m for all deformation
cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.44 The positive y end point modifications for s = 0.000898 m for all defor-
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
xiv
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
mation cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
B.1 The solution of the radius analysis and the linear regression of the anal-
ysis are nearly collinear. . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
xv
Acknowledgments
I am most grateful and indebted to my thesis advisor, Dr. William D. Mark, for
his generosity with guidance, patience, and encouragement that he has shown me here
at Penn State. I am especially indebted for the financial support which the Rotorcraft
Center has provided. I thank my other committee members, Drs. Gary Koopmann,
Stephen Hambric and Victor Sparrow, for their insightful commentary on my work. I
would also like to thank my wonderful wife, Sara, for her unyielding support over the
duration of this work. Her support has been essential to the completion of the techniques
contained herein.
1
Chapter 1
Introduction
From the inception of rotary machinery, gears have been manipulating and trans-
mitting power. From early wooden examples to modern involute drivetrains, gears have
been integral to the development of machinery and power manipulating technology. Most
modern gearing is conjugate [1], i.e. it is designed to transmit a constant rotational ve-
locity, and the involute tooth form [2] is the most common example of conjugate gearing.
Ideal involute gears transmit uniform rotational velocities without any error, but such
gears are not possible.
The displacement based exciter function known as the transmission error (T.E.)
[3, 4, 5, 6] is the accepted principal source for noise in involute gearing. In most gear
systems operating at speed, the load applied across the gear mesh dominates inertial
forces, and as such, the relative rotational position error of the meshing gears is directly
correlated to any vibration caused by the system. Excitation in the system is also
dependent upon the transmitted load. Put another way, an ideal gear pair would be able
to transmit a constant rotational velocity perfectly from the drive gear to the driven
gear under constant load conditions. For real gears, the difference in the position of
the output gear when compared to its ideal analog is the transmission error. This
transmission error can be directly traced to deviations from the perfect involute form
due to geometric differences and deformation under load.
There are two manifest requirements imposed upon a gearing system (tooth pro-
file) for transmission error fluctuations to be eliminated. One, the transmission error
must be constant through the range of the gear rotation for any mesh loading; and two,
the gear mesh must transmit a constant loading. These two fundamental quantities,
uniform force and velocity, govern the design of quiet gearing and bear on any design
consideration including stiffness, geometry or drivetrain layout.
The transmission error, as experienced in meshing gears, arises from two com-
ponents: geometric deviation from involute and deformations under load. Geometric
deviations can either be intentional or unintentional. Intentional deviations arise from
manufacturing modifications such as crowning or tip relief. Unintentional errors like
2
scalloping or improper finishing are also geometric deviations. Deformation across the
loaded tooth mesh also has two components. Both gross body compliance and Hertzian
compliance contribute to deformation under load. These deviations from the ideal in-
volute gear tooth surface for loaded gears are the primary cause of transmission error.
Eliminating or compensating for these deviation types can eliminate transmission error.
The purpose of this thesis is to describe a method for computing the optimal gear
and tooth design for the minimization of transmission error fluctuations and the mainte-
nance of constant transmitted gear mesh loading. This method is denoted as the Fourier
Null Matching Technique. Tooth geometry, tooth deformation, manufacturing error,
bearing stiffness and alignment must all be accounted for when tooth geometry is mod-
ified with the goal of reducing transmission error fluctuations for practical application.
The idealized case contained herein is the first step to a viable gear design.
The reduction of transmission error fluctuation requires the development of precise
compensatory gear geometry that, when loaded, accounts for all design and compliant
variations in the tooth-to-tooth meshing of a rotating gear pair. Effectively, material is
precisely added to the tooth face to exactly compensate for deformation under load. This
“zero sum” design should approach the performance of a rigid, ideal involute drivetrain
thereby transmitting a uniformly proportional rotational velocity, and thereby no trans-
mission error. In addition the total transmitted load across the gear mesh is held to be
constant. Careful control of the gear tooth geometry can extend this “zero sum” design
over a range of gear loadings. The final benefit of the Fourier Null Matching Technique
is that within the framework of maintaining constant transmitted load and transmis-
sion error, the frequency domain behavior of the meshing gear is such that the integer
bound tooth harmonics are eliminated by aligning the nulls of the frequency domain gear
tooth mesh behavior to the tooth meshing harmonics. The use of finite element analysis
(FEA), numerical optimization, linear algebra, and a variety of computational methods
are used in the solution of the “zero sum” gear design. The approach utilized herein is
believed to be novel.
3
Chapter 2
Previous Research
2.1 History of Gearing
Gears can be traced to the earliest machines. While the lever and wedge date
to the palaeolithic era [7], the other basic machines and engineering in the modern
sense can be historically traced to the Greeks [8]. Through surviving texts, Aristotle
and his followers are shown to have used and discussed the (gear-)wheel, the lever, the
(compound) pulley, the wedge and others. While limited by materials and analytical
techniques, simple machines and gears were used and continued to be developed and
considered by the Romans, the Arab world, and the Chinese [9]. From water clocks to
anchor hoists to catapults, the force-multiplying properties of gears were used by early
engineers throughout antiquity.
Though gears are simple in principle, their problems are far from trivial. In clocks
and windmills, gear wear was a significant problem and continued to be so throughout
the Renaissance. The mathematical and geometrical tools available to Renaissance en-
gineers were simply insufficient to solve the problem of the optimal gear tooth profile [8].
Leonhard Euler was the first to successfully attack the problem by showing that uniform
transfer of motion can be achieved by a conjugate and specifically, an involute profile [10].
While often ignored by his contemporaries due to being written in Latin and being highly
mathematical, Euler’s advances in planar kinematics and gearing in particular brought
modern gear analysis into existence.
While mechanical clocks drove the proliferation of gearing in the 15th and 16th
centuries, the industrial revolution brought an impetus to advance gearing in support of
the steam engine. By the end of the 18th century the École Polytechnique was established
in Paris and the modern academic study of kinematics and machinery had begun [9]. The
increased need for power manipulation on larger scales that accompanied the industrial
revolution brought about an explosion of fundamentally modern gearing.
Metallurgical, lubricative, and analytical advances continued through the 19th
and 20th centuries, but the form of gearing remained largely unchanged. Not until the
4
analysis of transmission error began in earnest in the mid 20th century did the form of
gearing deviate from Euler’s ideal profile.
2.2 Noise in Gearing
As a practical matter, audible gear noise resulting from transmission error is

subject to a torturous path from the gear mesh to the ear. From a meshing gear pair,
vibration must pass through bearings, shafts, gear cases, machine structures, mounts,
and panelling before any noise can be heard [11]. This type of path varies in specifics
but generally holds for any modern transmission. The cyclical nature of gear contact
(see Figure 2.1), and thereby, gear noise is expressed as cyclical deviation of rotational
position in the time domain (see Figure 2.2) and as a series of harmonics related to the
error cycles in the frequency domain (see Figure 2.3).
Fig. 2.1. An example of meshing gear teeth is shown. The global pressure angle, φ, the
base radius, Rb , and the pitch radius, R are indicated.
5
−3 Transmission Error
x 10
1.5
1
Transmission Error (mm)
0.5
−0.5
−1
100 200 300 400 500 600 700

Basal Distance, x (mm)
Fig. 2.2. An example transmission error profile is shown. The horizontal coordinate
can be thought of as roll distance as the gear rotates. The vertical coordinate is the
deviation from perfect transfer of rotational position expressed at the tooth surface. 1.2
cycles of gear rotation are shown.
Common measures used to address gear noise are isolation and damping through-
out the gear vibration path. However, the root of the problem lies in the gear tooth
interaction and any attempt at broadly reducing gear noise must focus on that cause.
By the late 1930’s, Henry Walker [12] was able to state that the transmission error
was the primary cause of gear noise. As gear analysis progressed in the 20th century,
the transmission error became the accepted cause of gear noise [3]. In the 1970’s and
1980’s the transmission error was shown to be directly proportional to audible gear
noise [11, 13]. Transmission error is now universally accepted to be the source of gear
noise [14, 15, 16, 17, 13].
6
Fourier Transmission Error Spectrum
−4
10
Fourier Magnitude (mm)
−5
10
−6
10
0 20 40 60 80 100 120 140

Gear Rotational Harmonic, n
Fig. 2.3. An example Fourier spectrum of the transmission error is shown. The horizon-
tal coordinate is the gear rotational harmonic. The vertical coordinate is the amplitude
of each rotational harmonic. The gear has 59 teeth.
Before rigorous analysis of involute gearing was available, gear noise was addressed
through rudimentary modification of the nominal involute tooth surface. This procedure
came to be known as crowning [18] or relief [19]. Tip relief is an effective tool to ensure
proper tooth clearance and nominal base plane action [14] which are critical to gear
noise [20]. In modern crowning, typically a single lead and a single profile modification are
combined and applied to the gear tooth surface. See the illustration in Figure 2.4. Axial
and profile crowning together [21, 22] have been shown to reduce gear noise in general and
address specific gearing issues such as misalignment. The addition of load computation,
mesh analysis, and line of contact constraints [23] have made the procedures’ results
robust and manufacturable.
Gear design for the minimization of gear noise, optimization of gear loading, and
maximization of manufacturability have been attempted throughout the field [24, 23, 19,
7
Fig. 2.4. Crowning: Typically a single lead modification and a single profile modifica-
tion, (a), linearly superimpose to give a crowned deviation from perfectly involute, (b).
Material, as described in this deviation from involute, is removed from the perfect invo-
lute tooth, (c), in a direction normal to the tooth surface thereby forming the crowned
tooth illustrated in (d).
8
25]. Some have attempted to isolate each individual component of gear noise and compu-
tationally superimpose these contributions to determine an optimal design [26, 27, 28, 29].
Others have used gear topology and curvature analysis to determine optimal results given
some installed constraints such as misalignment or mounting compliance [30]. These ef-
forts have typically resulted in some variation of a typical double crowned gear tooth
face with increased contact ratios [31], and these designs have been used in industry with
success. However, there are limitations to these analyses. First, full crowning reduces
the available contact area from a line to a point or a small ellipse [21]. This can result
in increased surface stresses relative to conventional gearing. Second, transmission error
and load transfer has not been kept constant [24, 22]. Third, frequency effects have been
largely ignored [27].
2.3 Current Gear Noise State of the Art
The dynamic action of installed gears has been shown to be effectively modelled by
a lumped parameter system where the drivetrain components are treated as individual
elements in the analysis [32]. These models tend to be constrained by classical gear
assumptions such as the plane of action for force transfer. There are, however, exceptions
to this [30]. The quality of the lumped parameter model is limited by the quality of the
transmission error input [33] which is, in turn, dependent on the gear tooth mesh model.
The components of gear mesh analysis have been shown to be highly dependent on gear
tooth compliance [26, 17] where the best compliance models take into account global
deformation, tooth deformation and contact mechanics [26, 34]. Among the current
methods for determining the static transmission error, all have their limitations and
many are reduced to two-dimensional or spur gear analysis [25].
Mark [5] has shown a complete and rigorous gear mesh solution for spur and
helical gears. Current noise optimization efforts are limited by the ability to model the
gear mesh, and the resultant data shows that dynamic response is not always directly
proportional to the static input. System resonances can alter and affect the dynamic
response. A full, rigorous mesh solution, however, may “zero” the input transmission
error and render the dynamic response moot.
2.4 Development of a Complete Solution
As early as 1929 it was known that tooth deformation was a major component of
practical gear design and analysis [35]. By 1938 the connection between deformation and
9
transmission error had been made by Henry Walker [12, 36, 37]. Walker noted that any
deflection made by a gear tooth under load acts the same as an error in the geometry of
the tooth. The result of this work began the use of tooth modifications to account for
the physics of tooth-to-tooth interaction under load.
By 1949 predictive methods were developed for the deflection of meshing gear
teeth [38, 39]. These methods were expanded and reduced to three primary phenom-
ena [40]: 1) cantilever beam deflection in the tooth, 2) deflection due to the root fillet
and gear body, and 3) local Hertzian [41] contact deformation. Also at this time, the
role of transmission error in vibration, noise, and geartrain performance was developed
by Harris [4, 3].
Due to the complicated and nonlinear nature of the deformation problem, com-
putational methods were necessary to further develop gear analysis. Conry and Seireg
[42] used a simplex-type algorithm to solve for deflection (or equivalently compliance)
from the three primary sources simultaneously. In their case, each source was calculated
based on classic methods similar to Weber [39]. Houser [14, 43] further developed these
methods with the work of Yakubek and Stegemiller [44, 45].
The finite element method (FEM) has proved to be the most robust method for
determining the compliance or deformation of loaded gear teeth [46, 47]. Two primary
considerations are present when applying finite element analysis (FEA) to gear tooth
compliance. First, the fine element meshes required to accurately model the gear body
and tooth cantilever deflections are computationally demanding. Second, gear deflection
models built around FEA often fail to account for Hertzian deformation. To address
this latter issue, Coy and Chao [48] developed a relation to govern the ratio (depth to
width) for the physical size of elements that allow for accurate prediction of local and
global deformation. This relation yields an element dimension given a load and another
element dimension as input. The third element dimension is collinear with the line of
contact and is unbounded.
Welker [49] took the work of Coy and Chao and showed that the relationship
between the transverse element dimensions for a given load is nonlinear. Welker showed
that there is a unique function relating element size to the size of the Hertzian contact
region. That relationship was expressed as a polynomial relation between the ratio of
element width to the Hertzian length (e/b) and the ratio of the Hertzian length to element
depth (b/c) as seen in Figure 2.5.
Jankowich [50] extended Welker’s work into three dimensions and Alulis [51] ap-
plied that work to a full gear body and helical gear tooth FEA model. Alulis used a
10
Fig. 2.5. From left to right: Given a finite element mesh of a gear tooth, any element
in that mesh that has an applied load also has size constraints. Any applied lineal load
along the length of the element creates a Hertzian contact region b and has a unique ratio
of element height c to width e associated with it that is required to accurately predict b.
Legendre polynomial representation of the compliance along a line of contact. The mesh
size and orientation was determined by Jankowich’s methods ensuring that the finite el-
ement mesh predicted the local deformation correctly. Alulis used 20 node isoparametric
brick elements in his finite element models.
Long before the work of Welker, Mark [5, 6] had developed a rigorous method for
describing the effects of geometry and compliance on transmission error. These methods,
coupled with the methods of Alulis, allow for a computational link to be developed
between transmission error, gear mesh loading, and tooth geometry. These are the tools
required to design gear teeth for the minimization of transmission error under constant
load.
2.5 Present Work
The present work demonstrates and implements a method for designing gear teeth
with a goal of minimizing transmission error fluctuations and load fluctuations. The
strong nonlinear interaction between the geometry and the resultant transmission error
11
requires careful constraining and bounding of any low transmission error solution. Signal
processing, optimization and Fourier analysis are used to guide the transmission error
minimization procedures. A very accurate tooth and gearbody stiffness model is required.
The gear tooth compliance functions are computed with Alulis’ [51] method for total gear
tooth stiffness. The open source Z88 finite element platform is used for the compliance
calculations. 20 node isoparametric brick elements are used to model the gear teeth.
12
Chapter 3
Preliminaries
A necessary prerequisite for understanding gear transmission error analysis is un-

derstanding of the fundamentals of gear geometry, gear mathematics and gear notation.
Definitions for all notation can be found in the appendices.
For a frame of reference, any tooth of interest on an analysis gear is usually
oriented at the top of the gear and the mating gear is located on top of the gear of
interest. See Figures 3.1 and 3.2.
Fig. 3.1. Standard gear diagram showing the analysis gear below and the mating gear
above it. The base plane projection of this helical gear pair is shown above the mating
gears. From Mark [5].
13
3.1 Gearing Constructs
The geometric profile for nearly all modern gears is based on the involute curve [52].
The involute curve, I, as seen in Figure 3.2 can be thought of as the arc the end of a
string traces as it unwraps from its spool with a radius equal to the base circle radius,
Rb . The roll angle, ², is the angle between the point of tangency, T , and the point
where the end of the string begins its unwrap as measured around the base circle. The
standard metrics for location on a gear tooth surface are the roll angle and the lead
location. The lead location is measured in the axial direction or out of the paper as
referenced by Figure 3.2. Continuing with the string analogy, the string length, T P 0 , is
identical to the arc length on the base circle for the angle ².
Using involute geometry, theoretically perfect gears would transfer a constant
rotational velocity without any deviation from that velocity. A deviation from that ideal
velocity transfer is the transmission error, which is a displacement form of excitation [53].
Perfect gears have an equivalent analog in a perfect belt drive running between two ideal
cylinders, and as such, all forces act in the plane of the conceptual belt between the two
cylinders. Ideal gears would mimic this characteristic exactly, and the involute curve is
the requisite geometry for maintaining this correlation. All gear tooth contact and all
tooth forces are in the plane of the analytical belt drive. This plane is named the base
plane and always intersects the horizontal (Figure 3.1) by the pressure angle, φ, in the
common frame of reference.
Continuing with the belt drive analogy, a meshing gear has a belt-cylinder de-
parture point at T (Figure 3.2), which is always located at the angle φ as measured on
the base plane from the pitch plane (Figure 3.4). As the gear rotates, the location of
T never varies in the global reference frame. See Figure 3.3.a. However, the analytical
Ti for any given point on the surface of the gear that is not in the plane of the two
gear axes is in a different location than the global tangency point, T . This analytical
Ti is measured in the local reference frame that is attached to the gear as shown in
Figure 3.3.b. These two reference frames are mathematically equivalent and constitute
a conceptual duality. This duality means that for a spur gear, a series of roll angles can
be thought of as a group on a motionless gear or as locations in the plane of the “belt”
as the tooth rotates. Furthermore, on a helical gear the same set of roll angles can be
conceptualized as positions along a line of contact.
There are four reference planes for gears (Figure 3.4). The axial plane contains
the two axes of the gear pair. The transverse plane is the plane that is perpendicular to
14
P’
R ∋
b
O
T φ
φi
Fig. 3.2. The involute curve, I, is traced by the end of a string unwrapped from a
cylinder (with radius Rb ). The roll angle, ², is the angle swept by the point of tangency,
T , for the string with the cylinder as the string unwraps. The pressure angle, φ, is the
angle of force for meshing gears. φ is measured against the pitch plane (the horizontal
plane perpendicular to the plane containing the gear axis). The construction pressure
angle, φi , is equal to φ when the “string” is unwrapped to the pitch point, P 0 .
15
(a) (b)
Fig. 3.3. Dual Construction: for a set of local pressure angles, φi , locations on a
single tooth profile or multiple profiles rotating through space can satisfy the resultant
geometry.
the gears’ axes. The pitch plane is perpendicular to both the transverse plane and the
axial plane and is located between the gear base cylinders. The final plane is the base
plane which is the plane of action for a meshing gear pair. The line of tooth contact
is always contained in the base plane. The global pressure angle, φ, is the angle of
intersection for the base plane and the pitch plane (See Dudley [54]).
For any involute helical or spur gear, there is an equivalent basic rack that would
mesh in an identical fashion to the actual gear. This rack can be thought of as an infinite-
radius gear with trapezoidal teeth or as a common gear that has been unwrapped to form
a flat, straight sequence of teeth. The line of contact for an involute helical gear tooth
is always in the plane of the equivalent rack tooth face and behaves precisely the same
as the rack tooth line of contact (Figure 3.5).
3.2 Mathematical Preliminaries
3.2.1 The Involute Curve
Most modern gearing begins with involute (black in Figure 3.6). With the involute
origin on the x axis, the involute equations are
x = Rb cos(²) + Rb ² sin(²) and y = Rb sin(²) − Rb ² cos(²). (3.1)
It should be noted that while most gearing analysis is done with a gear tooth upright
within the reference frame, the involute construction is nearly always built on the hori-
zontal axis as shown in Figure 3.6.
16
Fig. 3.4. The four gear planes: the transverse plane in blue, the pitch plane in green,
the axial plane in orange and the base plane in red. The mating gear would be located
directly above the gear cylinder for the gear of analysis shown in dark blue.
3.2.2 Angle Relations
The pressure angle, φ, and the pitch cylinder helix angle, ψ, are fundamental
angles of gear geometry. The pressure angle (see Figure 3.2) is the angle of force between
gear tooth working surfaces as measured against the pitch plane. The helix angle is the
angle of twist of the teeth as they wrap around a gear. Spur gears have straight teeth
and a helix angle equal to zero. All other relevant angles can be calculated from this
foundation. For example, the base helix angle can be found by
tan(ψb ) = cos(φ) tan(ψ). (3.2)

17
Fig. 3.5. A gear tooth (red) with its equivalent rack shown in blue and line of contact
in black. The green tooth profile intersects the line of contact at the pitch point.
18
Fig. 3.6. An involute curve (black) is commonly illustrated originating from the hori-
zontal or x-axis.
The normal pressure angle can be found by
tan(φn ) = cos(ψ) tan(φ). (3.3)
The intersection angle of the line of contact and the bottom edge of the equivalent
rack tooth in the rack tooth face plane is the load angle, θ0 (Figure 3.5). Geometric
analysis shows that the load angle can be found by
µ ¶
1 cos(φn ) 1
0 = − sin(ψ) sin(φ) . (3.4)
tan(θ ) cos(φ) sin(ψ) sin(φ)
Note that this is a corrected version of the same relation in Alulis [51].
19
θ′
φ
n φ
ψ
Fig. 3.7. θ0 , as measured on the face of the equivalent basic rack tooth, is the angle
between the line of contact (magenta) and the base of the rack tooth. The pressure
angle, φ, is measured in the transverse plane. As measured in the pitch plane, the pitch
cylinder helix angle, ψ, is measured between the tooth base normal and the transverse
plane. φn is the elevation of the tooth face normal.
Through the analysis that determines the local radius of curvature of the involute
gear tooth across the line of contact, a simpler and entirely equivalent relation was found
by
tan(θ0 ) = tan(ψ) sin(φn ) (3.5)
and is derived in Appendix B.

From Figure 3.2, the following angle relations,
tan(φi ) = ²i (3.6)
and
²i = θi + φi , (3.7)
are easily deduced. Another common measure for roll angle is the location spacing index,
βi . This is the arc angle swept between the center of the tooth and some location i on
the involute as measured from the center of the gear. βi together with ιi sum to φi (see
Figure 3.8).
20
βi
θ
i
ι
i
φ
i
Fig. 3.8. While the roll angle, ², can be broken up into its components, φi and θi , the
pressure angle, φi , can be further divided at the tooth thickness median into βi and ιi .
21
The circular pitch, p, gives an approximate tooth thickness by t ≤ p /2. The angle
between the tooth center and the pitch point is given by
π
βp = . (3.8)
2N
The preferred method to determine βi at any tooth face location is to use the pitch
spacing angle, βp . The subscript p denotes an angle as measured at the pitch point. The
sum of the spacing angle and the base cylinder arc length remain constant and can be
seen in
βi + θi = βp + θp . (3.9)
For any gear mesh contact, the forces are always in the base plane and are always
normal to the equivalent basic rack tooth face (see Figure 3.9). The force components
of F are found by
Fr = F sin(φn ),
Ft = F cos(φn ) cos(ψ), and (3.10)
Fa = F cos(φn ) sin(ψ).
and are broken into radial, transverse and axial components respectively as shown in
Shigley [52].
F
φ
n
φ
ψ Fr
Ft
F
a
Fig. 3.9. The force that is incident on a gear tooth can be broken into three components.
The helix and pressure angles are evident.
22
3.2.3 Gear Coordinates
There are two “projections” used when describing a location on a gear tooth face.
The x projection resides in the base plane as demonstrated in Figure 3.1. From this, the
two location variables are x, in the base plane and the axial coordinate normal to x, y.
Equivalently, a radial coordinate, z, may be used in lieu of x. The z projection is akin
to physically looking at the gear tooth face. z has a one to one correspondence to roll
angle.
The two most common gear coordinates are roll angle, ², and axial location, y.
Axial location, i.e. lead location, is measured as a lineal coordinate. The vertical or
radial coordinate, z, is defined by
z = Rb β² + C, (3.11)
where β = sin(φ) and C is some constant. It is useful to have the zero index of our
coordinates at the center of our gear tooth face. The center roll angle, ²0 , is defined as
the roll angle at the midpoint of the range, L in Figure 3.10. In this case, C = −Rb β²0
and z then becomes
z = Rb β(² − ²0 ). (3.12)
The range of z and y are contained within the lengths D and F respectively. Specifically,
−D/2 ≤ z ≤ D/2 and −F/2 ≤ y ≤ F/2. The face width, F , is the width of the tooth
face as measured in the axial direction i.e. the width of the gear. D is more complicated
and is defined below.
Since the line of contact always lies in the base plane, a single lineal coordinate,
s, can be distilled from z and y. For any given line of contact, there is a single s value
for each tooth. s is defined as
s = x − j∆, (3.13)
where j is tooth number and ∆ is base pitch. s is collinear with x as shown in Figure 3.1.
x, like s, is measured in the base plane. ∆ is the tooth spacing and is termed the base
pitch. x is the gear rotation distance and Θ is the gear’s global rotational position:
x = Rb Θ. (3.14)
23
z can be defined in terms of s and y:
z = βs + γy, (3.15)
where γ = ± sin(φ) tan(ψb ). γ is positive for a right-hand helix and negative for a
left-hand helix. It is often more useful to use Equation 3.15 in terms of s:
s = z/β − γy/β = Rb (² − ²0 ) ∓ tan(ψb )y. (3.16)
Like D and F for z and y, the range of s is bounded. In this case, the helix also
plays a role: −(F tan ψb +L)/2 ≤ s ≤ (F tan ψb +L)/2 which reduces to −L/2 ≤ s ≤ L/2
for spur gears. The upper bound of L is defined by the point at which the tooth face
exits the base plane. The lower bound of L is similarly defined by the mating gear. L
begins where the mating tooth face enters the base plane, and thereby comes in contact
with the analysis tooth face. See Appendix D and Figure 5 of [5].
L’
Fig. 3.10. Gear tooth ranges, L and D, relative to tooth rotation.

24
As is evident from Figure 3.10, D is defined through L:
D = Lβ = L sin(φ). (3.17)
L is a summation of the L contributions from each gear. For the lower gear (the gear of
analysis or gear (1)), L(1) is equal to the distance from the pitch point, P , to the point
where the addendum roll angle meets the base plane, ²a . L(2) is found in an identical
manner on the mating gear:
L = L(1) + L(2) = [Rb (²a − ²p )](1) + [Rb (²a − ²p )](2) . (3.18)
Parenthetical superscripts denote the gear with which each variable is associated.
3.2.4 Regular Gear Relations
The base circle radius can be found from the pressure angle and the pitch circle
radius. This same relation can used to find any radius, Ri , for some local pressure angle,
φi :
Rb
cos(φ) = . (3.19)
R
The base pitch, ∆, is the distance between the teeth on the base circle and can
be found by
2πRb
∆= . (3.20)
N
The contact ratios can be thought of as the average number of teeth in contact
in the axial and transverse planes. The axial contact ratio is defined in terms of the face
width and the helix angle. From
F tan(ψb )
Qa = , (3.21)
∆
it is apparent that the axial contact ratio is zero for spur gears. The transverse contact
ratio,
L
Qt = , (3.22)
∆
is simply the rotational contact length divided by the base pitch.
The standard working values for gearing are dependent upon the diametral pitch,
N N
P = = (3.23)
d 2R
25
and the circular pitch,

π
p= . (3.24)
P
R is the pitch circle radius and d is the pitch circle diameter. The diametral pitch, P ,
defines the addendum radius,
Ra = R + 1/P, (3.25)
and dedendum radius,

Rd = R − 1.25/P, (3.26)
in standard American Gear Manufacturers Association proportions. These radii are the
outer most radius of the gear and inner radius located between the teeth [52].
3.3 Transmission Error
There are two components of transmission error [5]:
ζ(x) = uj (x, y) + ηj (x, y). (3.27)
The first is due to geometric variations from perfect involute teeth, η. The second is due
to deformation as a result of tooth loading, u. ζ is the transmission error and j is tooth
number. Each of these values is a sum of components due to each of the two gears in
contact:
ζ(x) = ζ (1) (x) + ζ (2) (x), (3.28)

(1) (2)
uj (x, y) = uj (x, y) + uj (x, y), (3.29)
(1) (2)
ηj (x, y) = ηj (x, y) + ηj (x, y). (3.30)
The superscripts denote the gear associated with each term where (1) is the gear of
analysis and (2) is the mating gear. For u and η positive values denote “removal” of
material on the involute tooth face.
If u(x, y) is due to the load on a gear tooth, some load and some stiffness must be
brought into the analysis. Define Wj (x) as the total force transmitted by tooth pair j
as measured in the plane of contact. Given that KT j (x, y) is the local stiffness per unit
length of line of contact for tooth pair j, a relation of u can be developed by
Z
Wj (x) = KT j (x, y)uj (x, y)dl. (3.31)
LOC
26
dl is the differential portion of the line of contact and is measured in the base plane.
This differential length can be related to our orthogonal coordinates by
l = y sec ψb ∴ dl = sec(ψb )dy. (3.32)
By inserting the transmission error into the force equation and rearranging, the
transmission error in terms of loading and geometric deviations from involute can be
found [5]:
Z ³ ´
Wj (x) = KT j (x, y) ζ(x) − ηj (x, y) dl, (3.33)
LOC
ZyB ³ ´
Wj (x) = sec ψb KT j (x, y) ζ(x) − ηj (x, y) dy, (3.34)
yA
ZyB
KT j (x) = sec ψb KT j (x, y)dy, (3.35)
yA
ZyB
ηKj (x) = sec ψb KT j (x, y)ηj (x, y)dy, (3.36)
yA
X X X
W (x) = Wj (x) = ζ(x) KT j (x) − ηKj (x), ∴ (3.37)
j j j
P
W j ηKj (x)
η(x) = P +P . (3.38)
j KT j (x) j KT j (x)
The subscripts A and B denote the analysis endpoints on the line of contact.
3.4 The Tooth Force Equation
A very accurate method for determining the gear tooth local stiffness is developed
in detail by Alulis [51]. The gear tooth stiffness (or its inverse: the compliance) must
take into account both the global deformation including tooth cantilever deflection and
the local contact deformation. The following (Equations 3.39 through 3.50) summarizes
those techniques [5]. Each integral below is bounded by the portion of the line of contact
contained in the region of contact of the tooth face.
27
The total force transmitted by tooth pair j,

Z
Wj (x) = p0j (l, x)dl, (3.39)
can be treated as a function of the lineal force density, p0 , over the line of contact.
The deformation of the tooth can be calculated from this force per unit length
along the line of contact through a compliance influence function that includes bending,
shear, and Hertzian effects:
Z
u0j (l, x) = kj (l, l0 ; x)p0j (l0 , x)dl0 , (3.40)
where
u0j (l, x) = u0j (y sec ψb , x) = uj (y, x). (3.41)
kj (l, l0 ; x) is defined as the deformation at l on the line of contact on tooth pair j

due to a unit force applied at point l0 . l and l0 are reciprocal due to Maxwell’s Reciprocal
theorem [5]. Equation 3.40 is a representation of the Fredholm integral equation, and
the kernel notation used there is typical of integral equation usage. This equation can
be inverted to solve for the lineal force density,
Z
pj (l, x) = kj−1 (l, l0 ; x)u0j (l0 , x)dl0 . (3.42)
kj−1 (l, l0 ; x) is the stiffness influence function where the whole tooth local stiffness
is given by Equation 3.43.
Z
KT0 j (l, x) = kj−1 (l, l0 ; x)dl0 , (3.43)
where
KT0 j (l, x) = KT0 j (y sec ψb , x) = KT j (y, x). (3.44)
The tooth force equation,

Z Z
Wj (x) = kj−1 (l, l0 ; x)u0j (l, x)dl0 dl, (3.45)
is developed in terms of the tooth stiffness and can be expressed as

Z
Wj (x) = KT0 j (l, x)u0j (l, x)dl. (3.46)
28
In the familiar x and y terms the force equation is

Z
Wj (x) = sec ψb KT j (y, x)uj (y, x)dy. (3.47)
It now becomes critical to calculate the tooth stiffness. One must adjust the
inverse Fredholm integral equation for deformation under load (Equation 3.42) for a
constant deformation. This defines the load under constant deformation:
Z
pj (y, x) = kj−1 (l, l0 ; x)uj (x)dl0 . (3.48)
Using Equation 3.43 and solving gives the stiffness formulation,

Z
pj (y, x)
= kj−1 (l, l0 ; x)dl0 , (3.49)
uj (x)
where the full stiffness reduction becomes
pj (y, x)
KT j (y, x) = . (3.50)
uj (x)
As a practical matter, this representation of KT j is computed by linear algebra

for a series expansion of the loading and the lineal force density. Any function can be
represented as a series summation of orthogonal functions, and Legendre polynomials,
1 dn ³ 2 ´n
Pn (x) = n x − 1 , (3.51)
2 n! dxn
are one such family of orthogonal functions.

Modified Legendre polynomials are more appropriate for this formulation and are
defined by √
√ 2n + 1 dn ³ 2 ´n
Qn (x) = 2n + 1Pn (x) = x − 1 . (3.52)
2n n! dxn
The expansion for any function of x in modified Legendre polynomials is expressed
by
∞
X
f (x) = an Qn (x) (3.53)
n=0
where
Z1
1
an (x) = f (x)Qn (x)dx. (3.54)
2
−1
29
Expressed as a matrix equation, Equation 3.40 becomes
[an ]D = [C] [an ]L (3.55)
where each element along the line of loading in the finite element model generates three
modified Legendre coefficients. The subscript D denotes displacement and the subscript
L denotes loading.
In Equation 3.55 the displacement vector is the result of the loading vector acting
on the compliance matrix. Each term in the compliance matrix can be found through
applying the individual loading terms to the finite element representation of the gear
tooth, finding the resultant displacement and solving for the individual compliance terms.
See Alulis [51] for a detailed explanation. The full matrices are
   
a0,1 a0,1
    
 a1,1  c11 c12 c13 c14 c15 . . .  a1,1 
   
    
 a2,1   c21 c22 c23 c24 c25 . . .  a2,1 
    
    
 a0,2   c31 c32 c33 c34 c35 . . .  a0,2 
  =   (3.56)
 a1,2   c  a1,2 
   41 c42 c43 c44 c45 . . .  
    
 a2,2   c  a2,2 
   51 c52 c53 c54 c55 . . .  
  .. .. .. .. .. ..  
 a0,3  . . . . . .  a0,3 
   
.. ..
. .
D L
where in ap,e the first subscript is the polynomial order and the second subscript is the
element number along the line of contact.
Upon the completion of the full compliance matrix, a displacement can be found
for any loading or a loading can be found for a displacement. This allows for the com-
pletion of the calculation of the tooth loading force equation 3.47 or the transmission
error equation 3.38.
30
Chapter 4
The Main Procedure
4.1 Bounding the Problem
It has been established in previous chapters that the transmission error is the
exciter function for vibration and noise in meshing gears. It follows that the reduction of
transmission error is necessary in the minimization of noise. To eliminate transmission
error, the velocities and forces in a meshing gear pair must be constant throughout the
meshing of the gear pair. The zero transmission error requirement necessitates that for
a constant load a constant rotational velocity will result. Likewise, a constant load must
be transmitted in a meshing gear pair to eliminate the transmission error fluctuations.
Fourier analysis becomes important in dealing with transmission error because
it is useful to deal with the transmission error in the frequency domain. Fourier stated
that it is equivalent to describe a time-based function as a sum of frequency components.
The Fourier transform,
Z∞
X(f ) = x(t)e−i2πf t dt, (4.1)
−∞
allows a function to be analyzed in the frequency domain and to move back and forth
to the time domain with its inverse,
Z∞
x(t) = X(f )ei2πf t df. (4.2)
−∞
It is standard notation to denote Fourier frequency domain representations with capitol

letters and time domain function with lowercase letters.
For the computational methods included here, the discrete Fourier transform is
more appropriate:
N −1
1 X
Xn = xk e−i2πkn/N ; n = 0, . . . , N − 1. (4.3)
N
k=0
31
The discrete Fourier transform also has an inverse:
NX
−1
xk = Xn ei2πkn/N ; k = 0, . . . , N − 1. (4.4)
n=0
The complex Fourier series of the transmission error represents the behavior of
gear noise. The fundamental tooth-meshing harmonic frequency of n is 1/∆ meaning
that there is one Fourier coefficient per harmonic of tooth-to-tooth contact (Figure 4.1).
This is shown by
∆/2
Z
1
αn = ζ(s)e−i2πns/∆ ds. (4.5)
∆
−∆/2
The Fourier Uncertainty Principle becomes important in our ability to limit the
transmission error of a gear system. The Uncertainty Principle is most famous for
describing the inability to know both the location and velocity of an electron at the
same time. This idea has a counterpart in Fourier analysis. The more localized or
shorter in duration a function of time is, the more broad its frequency distribution.
Similarly, the more tightly the frequency components of a signal are clustered, the more
broad the temporal profile. There is a tradeoff between the compaction of the temporal
and frequency components of a function. Kammler [55] bounds the uncertainty with
1
∆s∆t ≥ . (4.6)
4π
This bears on the reduction of noise in gears in that for a function of finite duration
there can be no “zeroing” of the frequency components for all frequencies. For gears,
each tooth is in contact for a finite duration, therefore, the Uncertainty Principle directly
applies. However, one can eliminate the frequency component for a single frequency or
integer series of frequencies, e.g. f, 2f, 3f , etc. To minimize the noise in gears we
“zero” the harmonics of the tooth meshing fundamental (the frequency of tooth-to-tooth
contact) while maximizing the falloff of the remaining frequency components. Side bands
of the tooth meshing harmonics, which are caused by shaft rate imbalances, shaft rate
misalignment, etc., are minimized as well as these are also multiples of the tooth meshing
fundamental.
32
Fourier Transmission Error Spectrum
0.0001
1e−005
1e−006
1e−007
1e−008
0 20 40 60 80 100
Gear Rotational Harmonic, n
Fig. 4.1. Frequency domain theoretical gear noise is bound to integer multiples of the
primary gear rotational harmonic (n = 1). A vast majority of the noise power in real
world gears is found here as well.
4.2 The Role of Convolution in Gear Noise
Gear noise is mostly concentrated in the harmonics of the tooth meshing funda-
mental. The period of the fundamental is that of the adjacent tooth spacing, and to
illustrate this characteristic, a square function is used (Figure 4.2). The Fourier trans-
form of a square function is a continuous function with periodic zeros (or nulls). By
scaling the time domain transmission error so that the tooth meshing harmonics line up
with the zeros, a square wave type transmission error reduces the noise produced by the
gear substantially.
This central point should be emphasized. The total gear mesh force (and thereby
the transmission error) is designed in such a way that in the frequency spectrum, the
33
zeros in the force function directly overlap the tooth mesh harmonics and nullify the
primary component of gear noise. Subsequently, this technique shall be referred to as
the Fourier Null Matching technique.
(a) (c)
1 1
0.8 0.8
0.6 0.6
h(x)
h(x)
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 0 1 2 3 4
x x
(b) (d)
0 0
10 10
−1 −1
10 10
−2 −2
10 10
H(ω)
H(ω)
−3 −3
10 10
−4 −4
10 10
0 5 10 15 20 25 0 5 10 15 20 25
f f
Fig. 4.2. A unit square function (a) has a frequency spectrum (b) with zeros at integer
harmonics. A unit triangle function (c) which is the convolution of two square functions
also has a frequency spectrum (d) with zeros at the same integer harmonics. Note the
difference in the falloff of the two spectra.
A property of convolution that is utilized here states that given a function with
zeros in its Fourier transform, any convolution of that function also has zeros at the
same locations. It is established that the square function has zeros at integer frequency
34
values. The triangle function, which is the self convolution of a square function, also has
zeros at integer frequency values of its Fourier transform. A chain of convolutions can
create a continually changing profile in the time domain while the zeros of the frequency
domain are maintained. See Figure 4.3. The integral definition of convolution is
Z
f3 (s) = f1 (x)f2 (s − x)dx. (4.7)
(a)
0.8
0.6
h(x)
0.4
0.2
0 0.5 1 1.5 2 2.5 3 3.5 4

x
(b)
0.8
0.6
h(x)
0.4
0.2
0 0.5 1 1.5 2 2.5 3 3.5 4

x
Fig. 4.3. A narrow square function is convolved with a triangle function to create a
new rounded triangular function.
35
The second phase of effective noise reduction requires maximizing the falloff of the
non-integer frequency values for increasing frequency. Figure 4.4 illustrates the rate of
falloff of the Fourier transform as a function of the order of the convolution. For example,
a unit square function of width 10 convolved with a unit square function of width 2 will
result in Figure 4.4.a. This “second order” function results in a Fourier transform that
falls off as 1/f 2 . If Figure 4.4.a is convolved again by a unit square function of unit
width, Figure 4.4.c is the result. This “third order” function has a Fourier transform
that falls off as 1/f 3 .
(a) (c)
1 1
0.8 0.8
0.6 0.6
h(x)
h(x)
0.4 0.4
0.2 0.2
0 0
0 5 10 15 0 1 2 3 4 5 6 7 8 9 10111213141516
x x
(b) (d)
H(f) H(f)
−1 −1
f f
−2 −2
f f
−3 −3
f 0 f
10
0 10
−5
−5 10
10
0 2 0 2
10 10 10 10
frequency frequency
Fig. 4.4. The convolution of two square functions can produce a triangle function or a
trapezoidal function (a). Either is a “second order” function that has a Fourier transform
with asymptotic falloff of f −2 in frequency (b). A “third order” convolved function (c)
will produce a Fourier asymptotic falloff on the order of f −3 (d).
36
4.3 Convolution’s Relation to Gear Geometry
(a)
Α2
Α1
f(s)
0 ∋ −∆ 0 ∆
s
(b)
Α
f(s)
−∆−∋ /2 −∆+∋ /2 −∋ /2 0 ∋ /2 ∆− ∋/2 ∆+ ∋/2

s
Fig. 4.5. A square function of width ² convolves with a triangular function of width 2∆
to form a new function with a third order Fourier asymptotic falloff.
Given a triangle function, the convolution of a smaller square function produces

a rounded triangle of width 2∆ + ² as can be seen in Figure 4.5. The Fourier transform
of this function has zeros on integer multiples of 1/∆ and integer multiples of 1/². If
37
the triangle function represents the stiffness of a gear tooth and ∆ is the tooth spacing,
there are zeros on each of the tooth harmonics.
The curved triangle in Figure 4.5.b can be represented by a set of 5 equations
(4.8a-4.8e), where A = A1 A2 :
A
f1 (s) = (²/2 + ∆ + s)2 , −∆ − ²/2 ≤ s ≤ −∆ + ²/2; (4.8a)
2∆
A²
f2 (s) = (∆ + s) , −∆ + ²/2 ≤ s ≤ −²/2; (4.8b)
µ ∆ ¶
−1 ³ 2 2
´
f3 (s) = A ² /4 + s + ² , −²/2 ≤ s ≤ ²/2; (4.8c)
∆
A²
f4 (s) = (∆ − s) , ²/2 ≤ s ≤ ∆ − ²/2; (4.8d)
∆
A
f5 (s) = (²/2 + ∆ − s)2 , ∆ − ²/2 ≤ s ≤ ∆ + ²/2. (4.8e)
2∆
4.4 Developing the Procedure
Equations 4.8 can be linked to gear tooth parameters by the following relations:
u0
² = ∆α (4.9)
H
and  Ã !1+k 
HΩ  αu0 
A= 1+β (4.10)
∆ H
Here α is the force weighting factor, β is the load weighting factor, and k is the load slope
factor. They are assumed to be 1, 32/63, and 1/4 respectively. α allows for applying the
process to a loading scale that is less that the full design loading for the gear. β and k are
chosen to ensure that the gear tooth geometry behaves in a manufacturable way. u0 is the
tooth surface deformation and H is the maximum tooth surface deformation (assumed
to be 25 µm which is approximately the deformation experienced by fully loaded steel
gears). ∆ is the base pitch as described in Equation 3.20. Ω is defined as the force per
unit depth tooth surface deflection over a line of contact.
38
Assuming the loading function, Wj (s), from Equations 3.47 is equal to f (s) in
Equation 4.8c, the case where s = 0 yields a simple relation:
Ã ! Ã ! 
αu0 αu0 1+k
Wj (0) = αΩu0 1− 1 + β . (4.11)
4H H
The more general case of the tooth loading function for any s can be found by
combining Equations 4.8 through 4.10:
 Ã !1+k  Ã !2
ΩH  αu0  1+ α u 0 s
Wj (s) = 1+β + , (4.12a)
2 H 2H ∆
Ã ! Ã !
α u0 s α u0
−1 − ≤ ≤ −1 + ;
2H ∆ 2H
 Ã !1+k 
αu 0 ³ ´
Wj (s) = αΩu0 1 + β  1+ s , (4.12b)
H ∆
Ã ! Ã !
α u0 s α u0
−1 + ≤ ≤ − ;
2H ∆ 2H
 Ã !1+k   Ã !2 
αu 0 0 0 ³ ´
  αu − 1 αu s  2
Wj (s) = ΩH 1 + β − , (4.12c)
H H 4 H ∆
Ã ! Ã !
α u0 s α u0
− ≤ ≤ ;
2H ∆ 2H
 Ã !1+k 
αu 0 ³ ´
Wj (s) = αΩu0 1 + β  1− s , (4.12d)
H ∆
Ã ! Ã !
α u0 s α u0
≤ ≤ 1− ;
2H ∆ 2H
 Ã !1+k  Ã !2
ΩH  αu 0 α u 0 s
Wj (s) = 1+β  1+ − , (4.12e)
2 H 2H ∆
Ã ! Ã !
α u0 s α u0
1− ≤ ≤ 1+ .
2H ∆ 2H
It is important to note that the tooth loading is a symmetric function, i.e. Wj (s) =
Wj (−s). Equation set 4.12 gives a relation for the total applied load for a given line of
39
contact at a given surface deflection. The overall approach is to determine modifications

(material removal) from tooth surfaces that will provide constant transmitted loading, as
described above, and zero transmission error fluctuations (constant u0 for any prescribed
loading).
4.5 Applying Fourier Principles
At low loading levels, nominally perfect, rigid helical gears would possess a rect-
angular contact region. This ideal contact region could cover the majority of the tooth
surface or could be a smaller portion of the available area due to tooth geometry modi-
fications such as crowning. Also, as a rough approximation, the load transmitted across
the gear mesh for given surface deflection is proportional to the width of the line of
contact. Therefore, as the gear of analysis rotates, the load function (or alternately, the
load carrying capacity for a given line of contact) behaves in a manner that is nearly
proportional to the length of the line of contact.
If the axial and transverse contact ratios (Equations 3.21 and 3.22) are equal, the
s = 0 line of contact intersects the corners of this rectangular region for helical gears.
This is demonstrated in Figure 4.6. This configuration translates to the “triangular”
loading function shown in Figure 4.2.c. As a practical matter, the total stiffness of the
gear surface modifies this ideal contact region and the correct contact region must be
found.
Assuming a Qa = Qt = 1 initial contact region, the edges of the unmodified region
must exactly match the “triangular” profile for transmitted load throughout tooth-to-
tooth contact for all teeth. The light loading value shall be assumed to be the loading that
is required for a uniform tooth deflection of 1 µm. The “triangular” profile requirement
can be found by ¯ s ¯´
³
0 ¯ ¯
Wj (s) = αΩ0 u 1−¯ ¯ . (4.13)
∆
The force per unit depth can be found by
yB
Z (0)
αΩ0 = KT C (y, z)dl (4.14)
yA (0)
evaluated at s = 0.
40
3
s =∆
s = ∆/2
2 s=0 0.4
s = −∆/2
s = −∆
1
0.35
z (mm)
∋ (rad)
0
0.3
−1
0.25
−2
0.2
−3
−40 −30 −20 −10 0 10 20 30 40
y (mm)
Fig. 4.6. The black box denotes the available area on a tooth surface. The grey box
denotes the nominal Qa = Qt = 1 contact region. This region is only valid for very light
loading. As the line of contact travels from s = −∆ to s = ∆, the practical width of line
within the grey-boxed contact region behaves like a triangular function. The tooth root
is located at small ² and the tooth tip is located at the upper bound of ².
The solution to the initial contact region lies in determining the endpoints to the
load integral:
yZ
B (s)
Wj (s) = sec(ψb ) KT C (y, z)u0 dy. (4.15)
yA (s)
Note that u00 is a constant which leads to
yZ
B (s)
Wj (s) = u00 KT C (y, z)dl (4.16)
yA (s)
where the line of contact differential is defined as dl = sec(ψb )dy.

To bound the problem further, symmetry about the contact region should be
maintained. The change from the nominal rectangular contact region should be applied
in equal quantities to both ends of the line of contact. Assuming this symmetry, setting
Equation 4.16 equal to Equation 4.13 gives the endpoint adjustment, e, for the line of
contact at any s:
e = yA0 (s) − yA (s) = yB (s) − yB0 (s). (4.17)
41
Here the subscript 0 denotes the nominal rectangular region.

In this manner, the “light loading” region is determined and the appropriate
adjustments are made to the nominal rectangular region.
4.6 Extending the Contact Region
For any given s, the contact region solution for “light loading” should be extended
for a range of loadings. The intuitive fact that increased loading translates to an increased
range of contact provides the mechanism for maintaining reduced transmission error
fluctuations. The character of the extended lines of contact, under increasing load, are
controlled so as to maintain the traits of the convolution foundation of the Fourier Null
Matching Technique.
4.6.1 The First Step
Applying an increased loading to a line of contact across a tooth face will increase
the width and depth of that line of contact. The load equation, Equation 4.12, must be
expanded and evaluated. For example, the expansion of Equation 4.12c begins as follows
about s = 0:
Wj (s) = αΩu0 + . . . (4.18)
This allows two loading terms to be defined, the load mean, W , and the load variation,
ΣδWj , which contains all of the remaining terms in Equation 4.18. The result is written
as
Wj (s) = W (s) + ΣδWj (s). (4.19)
The load variation can be divided into two components which are developed from
the two ends of the line of contact:
ΣδWj (s) = δWjA (s) + δWjB (s). (4.20)
The contributions to the load variations at each end of the line of contact,
1
δWj (s) = KT C u01 θ1 , (4.21)
2
are resultant of the triangular regions at either end of the line of contact. See Figure 4.7.
42
u′0
Load Displacement
u′
1
yA−θ1 yA Tooth Face Lead Position yB yB+θ1
Fig. 4.7. The unmodified region of a line of contact, denoted by the flat area at loading
u00 , are extended on either side by some distance, θ1 . This distance corresponds to the
first design loading, u01 .
The extension to the line of contact, θ1 , must be isolated. Combining Equa-

tions 4.19 and 4.20 gives
Wj (s) − W (s) = ΣδWj (s) = δWjA (s) + δWjB (s). (4.22)
The equation is rearranged,
1³ A ´
Wj (s) − αΩu0 = KT C + KTBC u01 θ1 , (4.23)
2
and θ1 is isolated: ³ ´
2 Wj (s) − αΩu01
θ1 = ³ ´ . (4.24)
KTAC + KTBC u01
Here αΩu01 and W (s) are evaluated over the unmodified region. KTAC and KTBC
are evaluated at the geometrical mean of the triangular end regions of the line of contact.
4.6.2 Continuing the Expansion
For each line of contact on the tooth face, the previous sections describe the
solution of determining the “light loading” or unmodified region and the first step of
43
significant loading. To expand the range of valid gear loadings, the process must be
extended for additional steps of loading. See Figure 4.8.
0
u′0
Load Displacement
u′
1
u′2
θ2 θ1 Tooth Face Lead Position θ

1
θ2
Fig. 4.8. The line of contact includes two extensions beyond the unmodified contact
region. The first step extends the range by θ1 on either end and the second step extends
the previous line length by θ2 at each end.
Subsequent loading steps can be generalized, but follow a similar form as Equa-
tion 4.24. For the generalized case, a modification to Equation 4.19 must be made. For
any i > 0, Wj (s)i+1 can be found by the following:
f (s)
Wj (s)i+1 = W j i+1 + ΣδWj (s)i+1 . (4.25)
Much like Equation 4.15, the load contribution of the base region must be deter-
mined for case i + 1 by
Z
f (s)
W j i+1 = sec(ψb ) KT C (y, z)u0i+1 (y)dy. (4.26)
i+1
As there is no longer a uniform profile within the new load step, there is no
average load. Much as with the first load step, the load parts can be broken into the
extended regions for the new load step and the base region. Equation 4.21 must then be
44
generalized for θi+1 by
1³ A ´
δWjA (s) + δWjB (s) = KT C + KTBC δu0 θi+1 . (4.27)
2
θi+1 can be found by combining Equations 4.25 and 4.27 resulting in

³ ´
2 Wj (s)i+1 − Wf (s)
j i+1
θi+1 = ³ ´ . (4.28)
KTAC + KTBC δu0
f (s)
Wj (s)i+1 must be evaluated over the range including θi and W j i+1 should be
evaluated over the range including θi+1 . KT C should be evaluated at the endpoints of
the θi inclusive region.
4.7 Fourier Null Matching Technique
The Fourier Null Matching Technique can be summarized by reviewing the pri-
mary mechanisms of the process:
• The vast majority of transmission error arises from the tooth meshing harmonics
(Figures 4.9, 4.10).
• Tooth stiffness profiles can be created that have Fourier spectra with nulls on
integer multiples (Figure 4.11).
• These nulls can be aligned to the nominal tooth meshing harmonics (Figures 4.9,
4.11).
• The remaining gear rotational harmonics can be constrained by a decaying envelope

(Figure 4.12).
• The technique is valid over a range of loadings
The practical application of the transmission error analysis [5, 6] used in Figure 4.9
shows the predicted error spectrum. The predicted spectrum is comprised of tooth mesh-
ing harmonics. These harmonics are integer multiples of tooth-to-tooth contact events,
while measured transmission error spectra contain additional gear rotational harmonics
caused by tooth-to-tooth variations in tooth working surfaces caused by manufacturing
and installation errors [15, 14]. The Fourier Null Matching Technique holds the ana-
lytical transmission error constant while acknowledging that “real world” transmission
45
Analytical Fourier Transmission Error Spectrum

10−2
10−3
Fourier Magnitude (mm) 10−4
10−5
10−6
10−7
10−8
0 5 10 15 20 25
Tooth Harmonic, n
Fig. 4.9. An example Fourier transmission error spectrum is shown through the 25th
tooth harmonic.
error contains said additional gear rotational harmonic contributions (Figure 4.10). This
should be reiterated. The analytical transmission error is constrained to be constant by
the Fourier Null Matching Technique.
46
Illustrated Measured Transmission Error Spectrum

10−2
10−3
Fourier Magnitude (mm) 10−4
10−5
10−6
10−7
10−8
0 5 10 15 20 25
Tooth Harmonic, n
Fig. 4.10. An example continuous frequency noise spectrum is illustrated where the
noise energy is centered on integer multiples of the tooth meshing fundamental.
Example Continuous FNMT Transmission Error Spectrum

10−2
10−3
10−4
10−5
10−6
10−7
10−8
0 5 10 15 20 25
Tooth Harmonic, n
Fig. 4.11. The frequency domain envelope of a Fourier Null Matching Technique load
profile is shown.
47
Illustrated Composite Transmission Error Spectrum

10−2
10−3
10−4
10−5
10−6
10−7
10−8
0 5 10 15 20 25
Tooth Harmonic, n
Fig. 4.12. An illustration of a Fourier Null Matching Technique resultant transmission

error spectrum is shown.
48
4.8 Practical Work Flow
The algorithms behind the Fourier Null Matching Technique are the lifeblood of
the procedure and they demonstrate how to implement the above relations to determine
a quiet gear design. See Appendix C for code and finer detail.
Each line of contact is sequenced through each uniform deformation step. Two
index loops are used. Within the primary loop, the line of contact is indexed from s = 0
to s = −∆ and then from s = 0 to s = ∆ within the Qa = Qt = 1 bounding box.
See Figures 4.6 and 4.13. The nominal step of ∆/10 is used in the primary index loop.
Inside the primary loop, the secondary loop increases the uniform deformation from a
nominal 1 µm to a full load of 25 µm.
Fig. 4.13. An illustration of several lines of contact. The outer box represents the
available area on a tooth face. The inner box represents the Qa = Qt = 1 area which
bounds the line of contact for −∆ ≤ s ≤ ∆. The step size between lines of contact is
∆/4 in s while the horizontal coordinate is y and the vertical coordinate is z.
At each line of loading, starting with s = 0, a finite element model of the gear
is created. An estimate for the initial load is made. The element sizes on the line
of contact are found from the Zero Error Cylinder Relationship [51]. The compliance
curves for each finite element within the line of contact are determined. The compliance
49
formulation is then solved and the load for the constant deformation step is determined.
Using this load the iteration is repeated until the load converges. This gives an accurate
relation for the total gear compliance at the constant deformation in that step.
Once the nominal deformation and the associated loading is determined on the
line of contact at s = 0, the value of the total force per unit deformation over the full
line of contact is determined. This value, Ω0 , is the primary load capacity variable.
Continuing, the transmitted load in a gear mesh must remain constant. Therefore, Ω(s)
must maintain a triangular profile as measured in s.
Fig. 4.14. The lower triangular figure represents the force transmitted by a single tooth
pair as they come in and out of contact. More completely, the total force per unit
depth deformation over a line of contact as a function of s for each tooth-to-tooth cycle
is represented as this triangle over the range −∆ ≤ s ≤ ∆. The upper figure is the
superimposed transmitted forces of all tooth pairs. The total gear mesh transmitted
load (sum of the superimposed triangles) remains constant as the gear rotates.
Due to the compliance of the loaded gear tooth surface, the nominal, rectangular
Qa = Qt = 1 contact region cannot remain the final contact region. The final contact
region that corresponds to the target transmitted force will aberrate from that rectangle.
The endpoints of all lines of contact must be determined according to the target Ω(s)
profile. Once all lines of contact are determined, the bounds of the final contact region
is determined and consequently, the gear geometry for that loading is determined. An
illustration is shown in Figure 4.15.
50
Fig. 4.15. A conceptual illustration of the nominal “lightly loaded” contact region
(bounded in blue) that satisfies Equation 4.13. Note that the initial rectangular region
is not the bounds of the zone of contact that corresponds to a triangular transmitted
force profile. Several lines of contact are shown in red.
For the subsequent deformation steps, there is an extension of the line of contact
or contact region that is governed by Equations 4.24 or 4.28. An illustration of the first
extension loading step is shown in Figure 4.16.
In this manner, the entire sequence of deformation steps for all s for the entire
tooth face can be found, thereby determining the total tooth face geometry. This tooth
face solution possesses analytically eliminated transmission error fluctuations and loading
fluctuations. The resultant profile can be conceptualized as a either the addition of
material in the contact region, reducing region width as material is added or the removal
of material outside the contact region where more material is removed for light loading
cases and less is removed as loading and deformation are increased.
51
Fig. 4.16. A conceptual illustration of the first expanded contact region (narrow blue).
The contact region that corresponds to regular transmitted force profile (e.g. a triangle)
is not regularly shaped itself (e.g. not rectangular).
52
Chapter 5
The Results
5.1 Test Gear Parameters
The gear used to test the Fourier Null Matching Algorithm was determined prior
to the investigation. The parameters of the gear are shown in Table 5.1.
Table 5.1.
Test Design Gear
Variable Value
Number of Teeth, N 56
Face Width, F 60.96 mm
Pitch Radius, R 83.951 mm
Pitch Cylinder Helix Angle, ψ 18◦
Transverse Pressure Angle, φ 17.5◦
Young’s Modulus, E 207 GPa
Poisson’s Ratio, ν 0.3
There are a number of commonly used, but secondary, variables associated with
the test gear. These are listed in Table 5.2 for the design gear.
53
Table 5.2.
Test Design Gear Continued
Variable Value
Base Cylinder Helix Angle, ψb 17.217◦
Normal Pressure Angle, φn 16.692◦
Load Angle, θ0 5.3318◦
Base Cylinder Radius, Rb 80.065 mm
Diametral Pitch, P 0.33353 mm−1
Circular Pitch, p 9.4192 mm
Outer Radius, ROD 87.398 mm
Base Pitch, ∆ 8.9829 mm
Assumed Equal Basal Range, L 17.325 mm
Available Transverse Contact Ratio, Qt 1.9286
Available Axial Contact Ratio, Qa 2.1028
Operating Tooth Depth, D 5.210 mm
54
5.2 The Test Gear
A shaded representation of the gear can be seen in Figure 5.1. The computation
of the gear compliance is performed by finite elements on a single-tooth basis (see Fig-
ures 5.2 and 5.3). The FEA model is shown in Figures 5.4 and 5.5. Figure 5.5 shows
more clearly the concentration of elements surrounding the line of contact.
Fig. 5.1. A rendered example of the test gear.

55
Fig. 5.2. A closer view of the test gear showing the pitch point, P , and point of tangency,
T.
Fig. 5.3. A standard isometric view of the test gear. The pitch point, P , and point of
tangency, T are indicated.
56
Fig. 5.4. An example of a finite element model used for the compliance analysis.
57
Fig. 5.5. A close up of a finite element model used for the compliance analysis. This
model is for s = 0.8∆
5.3 Compliance and Loading
Compliance curves are critical to the calculation of an accurate tooth surface [51].
The iterative procedure for determining the correct finite element size and a loading
profile that are self-consistent is the process that drives the extreme computational load
of the Fourier Null Matching Technique. The iterative load profiles for the 1µm displace-
ment and each of the remaining five uniform displacements (5µm - 25µm) are shown (yel-
low, magenta, teal, red, green and blue in sequence). Note the strong proportionality to
load curves indicating the progressive loading behavior of a line of contact. Normalized
curves are indistinguishable from one another and are not shown. The loading curves
are summarized below. Line of contact locations indicated in Figures 5.6 to 5.24 can be
related to their locations on the tooth surface shown in Figures 5.26 to 5.31.
Note that the key marker for viability of the loading curves for constant deforma-
tion is the buttressing effect where the curve endpoints are raised relative to the middle
of the curve. This is necessary due to the unloaded surface outside of the curve. The
raised ends are compensatory for the loading change at the end points.
58
For line of contact s = −0.9∆

u′ = 25 µm
u′ = 20 µm
12000 u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
10000
8000
Normal Loading (N/m)
6000
4000
2000
0.0124 0.0126 0.0128 0.013 0.0132 0.0134 0.0136 0.0138

Axial Position (m)
Fig. 5.6. The loading curves for s = −0.008085 m

u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
12000 u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
10000
8000
6000
4000
2000
0
0.0105 0.011 0.0115 0.012 0.0125 0.013
Axial Position (m)

59

12000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
u′ = 10 µm
10000 u′ = 5 µm
u′ = 1 µm
8000
6000
4000
2000
0.0085 0.009 0.0095 0.01 0.0105 0.011 0.0115 0.012

Axial Position (m)

12000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
u′ = 10 µm
10000 u′ = 5 µm
u′ = 1 µm
8000
6000
4000
2000
0.0065 0.007 0.0075 0.008 0.0085 0.009 0.0095 0.01 0.0105 0.011
Axial Position (m)

60

12000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
u′ = 10 µm
10000 u′ = 5 µm
u′ = 1 µm
8000
6000
4000
2000
0.005 0.006 0.007 0.008 0.009 0.01

Axial Position (m)

u′ = 25 µm
u′ = 20 µm
12000 u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
10000
8000
6000
4000
2000
0.003 0.004 0.005 0.006 0.007 0.008 0.009

Axial Position (m)

61

u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
8000
u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
7000
6000
5000
4000
3000
2000
1000
0 0.002 0.004 0.006 0.008 0.01

Axial Position (m)

u′ = 25 µm
u′ = 20 µm
6000
u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
5000
4000
3000
2000
1000
−0.005 0 0.005 0.01

Axial Position (m)

62

u′ = 25 µm
u′ = 20 µm
4000 u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
3500 u′ = 1 µm
3000
2500
2000
1500
1000
500
−0.01 −0.005 0 0.005 0.01

Axial Position (m)
For line of contact s = 0

u′ = 25 µm
u′ = 20 µm
4000 u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
3500 u′ = 1 µm
3000
2500
2000
1500
1000
500
−0.01 −0.005 0 0.005 0.01

Axial Position (m)
Fig. 5.15. The loading curves for s = 0 m

63
For line of contact s = 0.1∆

u′ = 25 µm
4000 u′ = 20 µm
u′ = 15 µm
u′ = 10 µm
3500 u′ = 5 µm
u′ = 1 µm
3000
2500
2000
1500
1000
500
−0.01 −0.005 0 0.005 0.01

Axial Position (m)
Fig. 5.16. The loading curves for s = 0.00089833 m

u′ = 25 µm
5000 u′ = 20 µm
u′ = 15 µm
4500 u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
4000
3500
3000
2500
2000
1500
1000
500
−0.012 −0.01 −0.008 −0.006 −0.004 −0.002 0 0.002 0.004 0.006

Axial Position (m)

64
6000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
5000 u′ = 1 µm
4000
3000
2000
1000
−0.012 −0.01 −0.008 −0.006 −0.004 −0.002 0 0.002

Axial Position (m)

u′ = 25 µm
u′ = 20 µm
8000 u′ = 15 µm
u′ = 10 µm
u′ = 5 µm
7000 u′ = 1 µm
6000
5000
4000
3000
2000
1000
−0.011 −0.01 −0.009−0.008−0.007−0.006−0.005−0.004−0.003−0.002−0.001

Axial Position (m)

65

u′ = 25 µm
11000 u′ = 20 µm
u′ = 15 µm
10000 u′ = 10 µm
u′ = 5 µm
9000 u′ = 1 µm
8000
7000
6000
5000
4000
3000
2000
1000
−0.01 −0.009 −0.008 −0.007 −0.006 −0.005

Axial Position (m)

u′ = 25 µm
11000 u′ = 20 µm
u′ = 15 µm
10000 u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
9000
8000
7000
6000
5000
4000
3000
2000
1000
−0.011−0.0105 −0.01 −0.0095−0.009−0.0085−0.008−0.0075−0.007−0.0065

Axial Position (m)

66

u′ = 25 µm
u′ = 20 µm
10000
u′ = 15 µm
u′ = 10 µm
9000 u′ = 5 µm
u′ = 1 µm
8000
7000
6000
5000
4000
3000
2000
1000
−0.012 −0.0115 −0.011 −0.0105 −0.01 −0.0095 −0.009 −0.0085

Axial Position (m)
10000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
9000 u′ = 10 µm
u′ = 5 µm
u′ = 1 µm
8000
7000
6000
5000
4000
3000
2000
1000
−0.013 −0.0125 −0.012 −0.0115 −0.011 −0.0105 −0.01

Axial Position (m)

67
10000 u′ = 25 µm
u′ = 20 µm
u′ = 15 µm
9000 u′ = 10 µm
u′ = 5 µm
8000 u′ = 1 µm
Normal Loading (N/m) 7000
6000
5000
4000
3000
2000
1000
−0.014 −0.0138−0.0136−0.0134−0.0132 −0.013 −0.0128−0.0126−0.0124−0.0122

Axial Position (m)
The total force per unit deflection, αΩ, from the loading profiles is shown in
Table 5.3. The total loading and torque can be found from the average radius of the line
of contact and the deformation as shown in Equation 5.1.
τ = W R = αΩuR (5.1)
68
Table 5.3.
Summary of αΩ(N/µm).
s u = 1 µm u = 5 µm u = 10 µm u = 15 µm u = 20 µm u = 25 µm
−0.9∆ 0.2392 0.2332 0.2421 0.2426 0.2437 0.2443
−0.8∆ 0.4784 0.4843 0.5006 0.4908 0.5094 0.5144
−0.7∆ 0.7176 0.713 0.714 0.717 0.7198 0.7226
−0.6∆ 0.9568 0.9532 0.9539 0.9573 0.9602 0.963
−0.5∆ 1.196 1.206 1.211 1.21 1.214 1.217
−0.4∆ 1.435 1.436 1.453 1.455 1.462 1.467
−0.3∆ 1.674 1.672 1.672 1.674 1.676 1.677
−0.2∆ 1.914 1.912 1.912 1.916 1.913 1.914
−0.1∆ 2.153 2.152 2.154 2.116 2.092 2.152
0.0 2.392 2.391 2.357 2.336 2.317 2.303
0.1∆ 2.153 2.154 2.15 2.097 2.156 2.158
0.2∆ 1.914 1.92 1.92 1.921 1.922 1.923
0.3∆ 1.674 1.676 1.677 1.679 1.682 1.684
0.4∆ 1.435 1.433 1.433 1.434 1.435 1.437
0.5∆ 1.196 1.197 1.206 1.212 1.217 1.224
0.6∆ 0.9568 0.9518 0.9588 0.964 0.968 0.9748
0.7∆ 0.7176 0.7071 0.7126 0.7167 0.72 0.7262
0.8∆ 0.4784 0.4788 0.4828 0.4858 0.4882 0.4923
0.8∆ 0.2392 0.2399 0.2424 0.2441 0.2458 0.2492
69
For each deformation step, the total force per unit deflection should maintain a
nearly triangular profile. These profiles are shown normalized in Figure 5.25.
1
u′ = 1 µm
u′ = 5 µm
0.9 u′ = 10 µm
u′ = 15 µm
u′ = 20 µm
0.8
u′ = 25 µm
0.7
Normalized αΩ(s)
0.6
0.5
0.4
0.3
0.2
0.1
0
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
s/∆
Fig. 5.25. The nominal loading case results in a perfect triangle. The remaining load
steps deviate slightly from a triangular profile. Note that the rounded triangle is not
immediately apparent due to resolution in s.
70
5.4 Contact Region
Each of the loading curves in Figures 5.6 - 5.24 is for a specific line of loading.
Each line of loading is bounded by the Fourier Null Matching Technique. Figures 5.26
- 5.31 show the solution lines of contact for the test gear tooth. The endpoints of each
line of contact are shown in Table 5.4.
−3 u = 1e−006 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
1.5 s s s s s s s s s s
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
Fig. 5.26. The bounds of each line of contact and its position on the tooth face is shown
in the same orientation as Figure 4.6 for deformation step u = 1µm. Blue box denotes
Qa = Qt = 1. The lines of contact vary from s = −0.9∆ on the lower right to s = 0.9∆
on the upper left with a step between of ∆s = ∆/10. The tip of the tooth is positive z
and the root is negative z.
71
−3 u = 5e−006 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
Fig. 5.27. The bounds of each line of contact for deformation step u = 5µm. Blue box
denotes Qa = Qt = 1. The lines of contact vary from s = −0.9∆ on the lower right to
s = 0.9∆ on the upper left with a step between of ∆s = ∆/10. The tip of the tooth is
positive z and the root is negative z.
72
−3 u = 1e−005 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
73
−3 u = 1.5e−005 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
74
−3 u = 2e−005 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
75
−3 u = 2.5e−005 m
x 10
9∆ 8∆ 7∆ 6∆ 5∆ 4∆ 3∆ 2∆ 1∆
0. 0. 0. 0. 0. 0. 0. 0. 0. 0
= = = = = = = = = =
∆
.1
−0
=
s
∆
.2
−0
1 s
=
∆
.3
−0
=
s
∆
.4
0.5 −0
=
s
∆
.5
−0
z (m)
=
s
0 ∆
.6
−0
=
s
∆
.7
−0
=
s
−0.5 ∆
.8
−0
=
s
∆
.9
−0
=
s
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
y (m)
76
−3 Contact Region at Each Deformation Step

x 10
1.5
0.5
1µm
5µm
z (m)
0 10µm
15µm
20µm
25µm
−0.5
−1
−1.5
−0.015 −0.01 −0.005 0 0.005 0.01 0.015

y (m)
Fig. 5.32. The estimated contact region for the final geometry is seen. Note the well-
behaved portion of the tooth: negative s (lower right) for deformation steps from 5 µm
to 25 µm.
Table 5.4.
Summary of Line of Contact End Points. All values in m.
u = 1µm u = 5µm u = 10µm u = 15µm u = 20µm u = 25µm

s yBlower yBupper yBlower yBupper yBlower yBupper yBlower yBupper yBlower yBupper yBlower yBupper
-0.9∆ 0.012228 0.013863 0.012858 0.013233 0.012587 0.013504 0.012516 0.013574 0.012482 0.013609 0.012453 0.013637
-0.8∆ 0.010183 0.013008 0.011318 0.011873 0.010833 0.012359 0.010725 0.012467 0.010644 0.012548 0.010582 0.012609
-0.7∆ 0.0082266 0.012066 0.0099311 0.010362 0.0090778 0.011215 0.0088743 0.011418 0.0087872 0.011505 0.0087066 0.011586
-0.6∆ 0.0063243 0.011069 0.0081266 0.0092671 0.007199 0.010195 0.0069598 0.010434 0.0068595 0.010534 0.0067722 0.010621
-0.5∆ 0.0044272 0.010068 0.0060372 0.0084575 0.0052184 0.0092763 0.0049811 0.0095136 0.0048609 0.0096338 0.0047606 0.0097341
-0.4∆ 0.002596 0.0089998 0.0039457 0.0076501 0.0032398 0.008356 0.0030022 0.0085936 0.0028634 0.0087324 0.0027639 0.0088319
-0.3∆ -0.0015303 0.010227 -0.00002171 0.0087186 -0.0008551 0.0095519 -0.0011994 0.0098962 -0.0013931 0.01009 -0.0015084 0.010205
-0.2∆ -0.0057305 0.011528 -0.0044075 0.010205 -0.0052502 0.011048 -0.0056204 0.011418 -0.005886 0.011684 -0.0060804 0.011878
-0.1∆ -0.011914 0.014813 -0.011434 0.014333 -0.011892 0.014791 -0.012193 0.015092 -0.012415 0.015314 -0.012698 0.015597
0 -0.014495 0.014495 -0.014683 0.014683 -0.014765 0.014765 -0.014967 0.014967 -0.015175 0.015175 -0.015352 0.015352
0.1∆ -0.014998 0.0121 -0.014534 0.011635 -0.014982 0.012083 -0.015275 0.012376 -0.0156 0.012701 -0.015791 0.012892
0.2∆ -0.013797 0.0079994 -0.016132 0.010334 -0.014065 0.0082667 -0.013809 0.0080106 -0.014008 0.0082102 -0.014212 0.0084137
0.3∆ -0.012631 0.0039339 -0.014921 0.0062245 -0.014719 0.006022 -0.014975 0.0062785 -0.01279 0.004093 -0.01263 0.0039331
0.4∆ -0.011511 -0.00008526 -0.013277 0.0016815 -0.01308 0.0014844 -0.013234 0.001638 -0.011354 -0.00024167 -0.011272 -0.00032394
0.5∆ -0.010343 -0.0041521 -0.011644 -0.0028505 -0.011322 -0.0031727 -0.011365 -0.00313 -0.011523 -0.0029712 -0.011711 -0.0027835
0.6∆ -0.011356 -0.0060372 -0.012598 -0.0047956 -0.012292 -0.0051014 -0.012317 -0.0050762 -0.01245 -0.0049436 -0.013171 -0.0042229
0.7∆ -0.012315 -0.0079777 -0.013398 -0.0068946 -0.013132 -0.0071605 -0.013145 -0.0071478 -0.013894 -0.0063983 -0.013959 -0.0063334
0.8∆ -0.013253 -0.0099388 -0.014129 -0.0090624 -0.013909 -0.0092826 -0.014596 -0.0085958 -0.014625 -0.0085661 -0.014574 -0.0086177
0.9∆ -0.014033 -0.012058 -0.014524 -0.011567 -0.014908 -0.011183 -0.014887 -0.011204 -0.014813 -0.011278 -0.014766 -0.011324
77
78
5.5 Endpoint Modification
In each end point region of the progression of the Fourier Null Matching Tech-
nique, the variation and slope at the boundary of the contact region is the determining
factor for the utility and manufacturable of the gear as a whole. The end point solution
is the culmination of the work of Mark [5, 6] and Alulis [51] which results in a solution
for minimizing transmission error and loading variations and as a consequence, noise.
Figures 5.34 through 5.52 show the positive y side of the line of contact for each loading
set. The negative y side is identical since the problem is constrained to be symmetric
about the contact region. See the example in Figure 5.33. Each graph (and line of
contact at some s) is the solution for the full series of deformation cases at that location.
u′0
0
u′1
Load Displacement
u′
2
u′
3
u′
4
u′
5
θ5 θ θ3 θ θ Tooth Face Lead Position θθ θ θ θ
4 2 y1 yB1 2 3 4 5
A
Fig. 5.33. Each line of contact is constrained to have symmetric adjustments to the
length of the line of contact for each deformation step. The positive-y extension profile
is a mirror of the negative-y deformation profile.
79
−5
x 10 For line of contact s = −0.9∆
0
−0.5
u (m) −1
−1.5
−2
−2.5
−1.2 −1 −0.8 −0.6 −0.4 −0.2 0

y + Y (m) where Y = 0.0138626 −3
x 10
bu bu
Fig. 5.34. The positive y end point modifications for s = −0.00808 m for all deformation
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−2 −1.5 −1 −0.5 0
y + Ybu (m) where Ybu = 0.0130082 −3
x 10
cases.
80
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−3 −2.5 −2 −1.5 −1 −0.5 0

y+Y (m) where Y = 0.012066 −3
x 10
bu bu
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−3.5 −3 −2.5 −2 −1.5 −1 −0.5 0

y + Ybu (m) where Ybu = 0.0110694 −3
x 10
cases.
81
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−3 −2.5 −2 −1.5 −1 −0.5 0

y + Y (m) where Y = 0.0100675 −3
x 10
bu bu
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−2.5 −2 −1.5 −1 −0.5 0

y + Ybu (m) where Ybu = 0.00899975 −3
x 10
cases.
82
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−3 −2.5 −2 −1.5 −1 −0.5 0

y + Ybu (m) where Ybu = 0.0102272 −3
x 10
cases.
x 10
−5 For line of contact s = −0.2∆
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−2.5 −2 −1.5 −1 −0.5 0

y + Ybu (m) where Ybu = 0.011518 x 10
−3
cases.
83
x 10
−5 For line of contact s = −0.1∆
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−15 −10 −5 0 5
y + Ybu (m) where Ybu = 0.0148132 x 10
−4
Fig. 5.42. The positive y end point modifications for s = −0.000898 m for all deforma-
tion cases.
−5 For line of contact s = 0

x 10
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−8 −6 −4 −2 0 2 4 6 8
y + Ybu (m) where Ybu = 0.0144947 x 10
−4
Fig. 5.43. The positive y end point modifications for s = 0 m for all deformation cases.
84
−5
x 10 For line of contact s = 0.1∆
0
−0.5
u (m) −1
−1.5
−2
−2.5
−15 −10 −5 0 5
y + Ybu (m) where Ybu = 0.012147 −4
x 10
Fig. 5.44. The positive y end point modifications for s = 0.000898 m for all deformation
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

y + Ybu (m) where Ybu = 0.00803416 −3
x 10
cases.
85
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

y + Ybu (m) where Ybu = 0.00397132 x 10
−3
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5

y + Ybu (m) where Ybu = −7.62256e−005 x 10
−3
cases.
86
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−1 −0.5 0 0.5 1
y + Ybu (m) where Ybu = −0.00412936 −3
x 10
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−1.5 −1 −0.5 0 0.5 1 1.5

y + Ybu (m) where Ybu = −0.00603724 −3
x 10
cases.
87
−5
0
−0.5
u (m) −1
−1.5
−2
−2.5
−1.5 −1 −0.5 0 0.5 1 1.5

y + Ybu (m) where Ybu = −0.00797768 x 10
−3
cases.
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−1 −0.5 0 0.5 1
y + Ybu (m) where Ybu = −0.00993876 x 10
−3
cases.
88
−5
0
−0.5
−1
u (m)
−1.5
−2
−2.5
−8 −6 −4 −2 0 2 4 6 8
y + Ybu (m) where Ybu = −0.0120567 x 10
−4
cases.
89
5.6 Discussion of Computed Contact Regions
There are four key concepts that bear on the results in Figures 5.26 through 5.31.
First, the Hertzian stiffness for gear tooth contact increases with increased deformation.
Welker [49] shows these relations very succinctly in Appendix A. The Hertzian stiff-
ness is also very small for small deformations relative to total tooth stiffness. Second,
the Hertzian stiffness is proportional to the length of a line of contact to first order.
With this, the length of a line of contact can be used as a qualitative measure of the
Hertzian contribution to total stiffness. Third, the nonlinear increase in Hertzian stiff-
ness decreases with increasing deformation. Again see Welker, Appendix A. Finally, the
bending stiffness of the tooth deformations is larger near the root (negative s) and de-
creases as a line of contact travels toward the tip (positive s). Also note that the imposed
loading, as shown in Figure 5.25, is symmetric about s = 0 for each deformation case.
Figure 5.26 shows the assumed negligible loading case. This initial case demon-
strates the Hertzian stiffness behavior for small deformations. For these small deforma-
tions, the Hertzian stiffness is also small and as a result, almost all of the deformation for
this case is of the Hertzian variety. The bending stiffnesses are not operative because the
effective bending stiffnesses are much higher than Hertzian stiffnesses for light loading.
Without tooth bending or gross tooth stiffness as significant contributors to global tooth
stiffness, the computed bound lines of contact are largely symmetrical about s = 0 as a
result of the symmetrical imposed total loading.
The relative and qualitative magnitude of the bending stiffness contribution can
be seen in the differences in the negative s and positive s behavior in Figure 5.26.
The length for a given positive s line of contact is slightly longer than its negative s
counterpart. The difference in the lengths of these pairs of lines of contact is traced to
the increase in the Hertzian stiffness contribution to total stiffness that is required near
the tip of the gear tooth (positive s). In the tip region, the tooth bending stiffness is
smaller compared to that at the tooth root (negative s). Consequently, the positive s
Hertzian stiffness contribution must be larger than that of the negative s counterpart.
The length of the line of contact is therefore correspondingly longer in the positive s
portion of the tooth.
Figure 5.27 shows the lines of contact for the first deformation step that is designed
to be fully engaged. This deformation step is 5 µm. The distribution of the Hertzian and
the bending components of total stiffness has changed with respect to the nominal light
loading case shown in Figure 5.26. The applicable bending stiffness contribution to total
90
stiffness in the positive s portion of the contact region is much smaller than that in the
negative s portion. This extends the lines of contact. This effect is a result of the imposed
symmetric loading shown in Figure 5.25 and the analytical tooth behavior. Conversely,
the negative s portion of the contact region exhibits shorter lines of contact relative to
their positive counterparts. Here the Hertzian component of total stiffness is reduced,
which results in the reduced contact line lengths. It is obvious that such a reduction
relative to the nominal line of loading is not physically achievable. An increase in the
load weighting factor’s assumed value (β = 32/63) should help to address this physical
discrepancy.
Figure 5.28 shows that the subsequent deformation step to Figure 5.27 behaves
in a similar fashion. The bending stiffness portion of the total stiffness at the root (neg-
ative s) is relatively large allowing shorter lines of contact and the associated Hertzian
stiffness carry the majority of the imposed loading in this region. Since the Hertzian
component of stiffness is largely proportional to length of the line of contact the neg-
ative s contact lines are shorter by comparison. As the deformation and loading are
incrementally increased a nonlinear increase in total computed stiffness results from the
Fourier Null Matching Technique. Continuing in the negative s region, the load weight-
ing factor, β, is sufficiently large to ensure a positive increase in the length of the line
of contact as tooth deformation is increased. Figures 5.29, 5.30, and 5.31 show the
positive contact line extension for increasing deformation in the negative s region. The
load weighting factor, β, is developed to “overcome” the nonlinear increases in Hertzian
stiffness in Equations 4.12 and 4.28 in such a way that as deformation and loading are
increased the length of the line of contact is also increased.
In the positive s region, the bending stiffness is smaller than in the root region.
Hence, the Hertzian contribution to total stiffness is greater in proportion to the total
requisite stiffness as a result of the imposed loading. With this greater Hertzian com-
ponent, the length of the lines of contact in the positive s region must be greater as
well. This is clearly shown in Figures 5.28 through 5.31. The behavior of the larger
nonlinear Hertzian stiffness contribution demonstrates that the load weighting factor, β,
is not sufficiently large in these analyses to “overcome” the actual nonlinearities in the
Hertzian contribution. This is expressed as negative extension to the length of the line
of contact as deformation is increased by Equation 4.28. A value for the load weighting
factor, β, larger than the 32/63 used could have corrected this issue and the nonpractical
resultant design geometry. The positive s region demonstrates that for future analyses, a
larger load weighting factor, β, and possibly a modified load slope factor, k, is called for
91
to realize a tooth surface design that is physically feasible. The load weighting factor,
β, controls the nonlinear stiffness characteristic of the model which by Equation 4.28
controls the extension of the lengths of the lines of contact.
The analysis contained herein is for a single gear, and using the Fourier Null
Matching Technique for a practical application will require the analysis of a mating gear
pair. Careful control of the mating gear pair may allow for the joint mesh stiffness to
maintain good behavior throughout the gear rotation.
92
Chapter 6
Summary and Conclusions
6.1 The Fourier Null Matching Technique
The current state of the art in gear analysis has been expanded to create a proce-
dure for designing an optimal tooth face modification for reducing or eliminating trans-
mission error variation across a range of loadings. Despite the limitations of the positive
s contact region, the technique has shown that a rigorous gear mesh analysis protocol
can be used to design an optimal contact region for a helical gear for a given deformation.
At a given deformation step, the helical gear of analysis and its results show that
the contact region can be controlled in such a way as to create a mesh stiffness profile
that maintains a constant transmission error, maintains a constant transmitted load, and
creates operational nulls on the tooth meshing harmonics. This constitutes a significant
advance in the analytical state of the art for gearing.
It should be duly noted that the transmission error analysis that is the basis
of the Fourier Null Matching Technique is a static analysis. The transmission error
is accepted to be largely proportional to measured gear noise: implying that dynamic
effects are secondary [15, 11, 13]. Given that the transmission error is, in large part,
proportional to and the root of gear noise and vibration, the Fourier Null Matching
Technique significantly reduces the source of gear noise by eliminating the tooth meshing
harmonics of transmission error while maintaining constant transmitted load. The tooth
surface modification techniques contained herein expand the capabilities of standard gear
analysis, prediction and design.
The example computation shows that using a very accurate gear stiffness model,
it is feasible to compute the contact region on a gear tooth face that will provide a
constant transmission error and a constant mesh loading over a range of mesh loadings
and transmission errors.
93
6.2 The Practicality of the Tooth Modification
Gear grinding equipment can typically reach a geometric accuracy of approxi-

mately 1 µm [54]. With a mean step distance on the order of 0.5 mm, it is realistic
to attempt to manufacture the negative s portion of contact region for the deformation
ranges of 5 µm to 25 µm. The positive s portion of gear tooth surface is not manufac-
turable given the negative extension results. In addition, the nominal “light loading”
deformation (1 µm) does not sufficiently engage the tooth compliance to maintain the
Fourier Null Matching Technique. The first step of the analysis should be increased in
magnitude and reevaluated.
The ultimate practicality of the as-designed tooth face will likely lie in the joint
gear mesh solution’s ability to compensate for the expressed compliance behavior near
the tooth tip (positive s). Additional compromises for manufacturability, for the non-
ideal realities of installed gear sets, and for life cycle longevity may have to be made.
With the careful design of whole gear systems, the technical limitations of quiet gears
that are made apparent by the Fourier Null Matching Technique can be addressed in the
majority of powertrain applications.
The base outline of the contact region for any given deformation step is likely
to have significant applicability and noise reduction capability across a limited range of
loading levels. As a practical matter the contact regions demonstrated in Figures 5.27-
5.31 may be construed as a rough guideline for the form of the contact region that is
necessary for quiet operation. Within the limits of manufacturability, material knowl-
edge, and implementability, the Fourier Null Matching Technique has the potential to
significantly alter gear analysis, gear design and powertrain operation.
6.3 Limitations of the Analysis
The Fourier Null Matching Technique requires high precision and as such, there
are a number of limiting factors in the technique’s utility. First and foremost, the
technique is limited by the knowledge of the gear material. Lack of knowledge of grain
structure, Young’s modulus and Poisson’s ratio adds variability and uncertainty to the
full technique. Material properties are crucial to accurate finite element computation.
Hence, normal material property knowledge is limiting.
Implementing this method requires that material be removed from the tooth face.
In the case of the afore-described gear, at least 25 µm of material must be removed from
94
the tooth face outside of the contact region. The analysis gear has a pitch thickness of
approximately 4.7 mm. Removing the full design deformation amounts to a thinning of
more than 0.5%. Removing this material will modify the stiffness response of the gear
tooth and will require additional compliance iterations due to the change in physical
tooth properties, especially in the positive s region of the tooth surface.
Other factors that are present in normal gear and powertrain operation are ignored
in this analysis. Friction and lubrication properties are small and do not lay in the plane
of action of a meshing gear pair. In addition, the direction and mechanisms of friction
forces in are largely self-cancelling [14] in helical gears. Consequently, friction effects
are neglected here. However, Liu and Parker [56] have shown that the effects of friction
do express themselves in the dynamic response of the gear system. Friction should be
included in the full design of any specific installation configuration.
The final neglected situational mechanism is the structural environment of a
normal meshing gear pair. The bearings, shafts and supporting structures affect the
transmission error by introducing eccentricity and misalignment. These effects can be
accounted for systematically, but only if the magnitude of these effects are known ex-
plicitly. Sensitivity to misalignment and eccentricity should be statistically bounded for
further analysis and possible physical testing.
Shaft eccentricity affects gear noise by providing a once-per-revolution (gear rota-
tional harmonic n = 1) tooth-spacing error. Since few teeth are in contact at any given
instant, any eccentricity effect would be small.
Misalignment would affect the Fourier Null Matching Technique generated gear
much like a standard gear [57, 58]. Axially crowned gears are less affected by misalign-
ment than standard gears [59]. However, the effective contact ratio of axially crowned
gears are reduced compared to the Fourier Null Matching Technique gear or a standard
gear, and the corresponding noise properties will reflect that. As the unmodified region
maintains the involute profile without modification, misalignment will not greatly affect
vibration generation while the whole line of contact boundary is contributing to the
gear mesh. Misalignment will, however, alter the surface contact stresses. On one end
contact stresses will be increased while on the other end they will be reduced. Sufficient
misalignment for loss of contact will create additional gear noise much as a standard
gear will behave under such a scenario [58].
95
6.4 Extending the Process
This analysis is provided in a one-sided context, i.e. only the test gear is analyzed
without consideration of its mate. To the extent that complimentary gear parameters
are needed, any mating gear in the present analysis would be an ideal, rigid involute
gear of the same size as the gear of analysis. For an installed gear pair, the geometrical
compensation for the transmission error mechanisms must be applied to the gears in
tandem. Each gear must share in the surface modifications and the total gear pair mesh
stiffness must be evaluated through the Fourier Null Matching Technique.
The computation should be expanded to determine the sensitivity of the method
to the parameters β and k. Other values of these constants may lead to more gradual
slopes with respect to increasing uniform elastic deformation as noted above. This pro-
cess would require significant additional effort due to the intense computational load of
the technique. In addition, the sensitivity of the Legendre polynomial representation of
the compliance loading curves to the number and type of axial elements must be de-
termined. The dependence on the axial element configuration and number to the final
evaluation of the load integral, Equation 4.15, must be parameterized.
In using the Fourier Null Matching Technique for functional gear design, certain
modifications could be made to address other real-word issues. For example, misalign-
ment can be addressed by introducing a curvature to the gear tooth face that allows the
gear to behave similarly in both the misaligned and the correctly mounted position. In
the Fourier Null Matching Technique, the step functions used in each deformation step
could be altered to allow for curvature down the line of contact. In this manner, mis-
alignment can be addressed within the context of the Fourier Null Matching Technique.
The next conceptual step in the development of the Fourier Null Matching Tech-
nique should be to replace Alulis’ computationally intensive compliance formulation with
a more efficient model. Ideally this future compliance model will be computationally ef-
ficient enough to run statistical variations and global optimizations on the Fourier Null
Matching Technique. An elegant yet accurate relation for tooth compliance would be of
great benefit to any extension of the present analysis for the entire gear industry. With
such an advance, wide scale adoption and implementation of the Fourier Null Matching
Technique within the gear industry may ultimately be feasible.
96
6.5 Closure
The goal of this dissertation is to answer a singular question: Can a set of mod-
ifications for an involute tooth working surface be devised such that the resultant gear
will transmit a constant load and zero transmission error fluctuation for a non-negligible
range of loadings? Contained herein is a procedure that utilizes the nonlinear tooth stiff-
ness properties of a gear tooth, including all Hertzian contributions, to attempt that goal.
The results of this procedure, while being a mere first step, answer this dissertation’s
central query in the affirmative.
97
Appendix A
Notation
In gearing literature, the notation is loose and rarely self consistent. The notation
listed in Table A.1 is a compromise of the most common usage and is used throughout
this document.
Table A.1: Gearing Notation
Variable Comments
I The involute curve
² Roll angle for any point, i, on I
²p The roll angle for the pitch point, P
²0 The roll angle at the center of the basal range, L
φ The global pressure angle, equal to φi at i = P
φi The pressure angle for any point, i, on I
φn The normal pressure angle
θ The pitch point involute base cylinder arclength, ²p − φ
θi The general involute base cylinder arclength, ² − φi
O The involute origin
T The conceptual involute tangency point
P0 The pitch point
R The pitch circle radius
Rb The base circle radius
ψ The pitch cylinder helix angle
ψb The base cylinder helix angle
N The number of teeth on a gear
∆ The base pitch
Qa The operating axial contact ratio
Qt The operating transverse contact ratio
F The axial face width
Continued on Next Page. . .
98
Table A.1 – Continued

Variable Comments
D The operating tooth depth
L The basal tooth contact range
y The axial tooth face coordinate
z The radial tooth face coordinate
x The basal tooth coordinate
s The local basal tooth coordinate
P The diametral pitch
p The circular pitch
d The pitch diameter
β sin(φ)
γ ±β tan(ψb ): positive for right hand helix
βi The angle measured from the center of the tooth to i on the involute
ιi φi − βi
θ0 The load angle of the line of contact
Rd The dedendum radius
Ra The addendum radius
²a The roll angle at the addendum
t The tooth thickness
The transmission error notation contained within this document is derived from
Mark [5]. As with all of the notation throughout this research, there are small variations
for the purpose of self-consistency. See Table A.2.
The optimization work in this thesis uses additional notation that is listed in
Table A.3.
99
Table A.2.
Summary of Transmission Error Notation
Variable Comments
ζ The transmission error
u Tooth face deformation due to loading under the tooth mesh
η Tooth face variation due to geometry
j Tooth number
Table A.3.
Summary of Optimization Notation
Variable Comments
A The convolution product amplitude
∆ The length of the linearly increasing portion of the convolution precursor
² The width of the square convolution precursor
u0i Iteration, i, of constant deflection
H The maximum design deformation (25 µm in this analysis)
Ω The force per unit depth u0 evaluated over the line of contact
α Force weighting factor (assumed to be 1)
β Load weighting factor (assumed to be 32/63)
k Load slope factor (assumed to be 1/4)
u The deflection of a gear tooth surface on a line of contact
p The lineal force density along the line of contact
kj The tooth compliance influence function
KT j The total stiffness of the gear tooth face
Wj The loading force for a given line of contact between meshing teeth
100
Appendix B
Derivations
B.1 Simplifying the Load Angle Equation
Equation 3.4 can be rearranged to the Equation B.1.
cos(φ) sin(ψ) sin(φ)

tan(θ0 ) = (B.1)
cos(φn ) 1 − sin2 (ψ) sin2 (φ)
Multiplying top and bottom by cos2 (ψ) cos2 (φ) sin(φn ) and simplifying produces
Equation B.2.
tan(ψ) tan(φ) tan(φn ) 1

tan(θ0 ) = (B.2)
sin(φn ) cos(ψ) sec (ψ) sec (φ) − tan2 (ψ) tan2 (φ)
2 2
Substituting Equation 3.3 and again multiplying top and bottom by sin(φn ) fur-
ther simplifies the equality when the Pythagorean trigonometric identity is used in the
denominator.
tan2 (φ) csc2 (φn )

tan(θ0 ) = tan(ψ) sin(φn ) (B.3)
(tan2 (ψ) + 1)(tan2 (φ) + 1) − tan2 (ψ) tan2 (φ)
Again using the Pythagorean trigonometric identity and expanding the denomi-
nator produces Equation B.4.
tan2 (φ)(cot2 (φn ) + 1)

tan2 (φ) + tan2 (ψ) + 1
Expanding the numerator and applying Equation 3.3 reduces the numerator. Us-
ing the Pythagorean identity a final time reduces the fraction to unity.
tan2 (φ) + sec2 (ψ)

tan2 (φ) + tan2 (ψ) + 1

101
B.2 Radius of Curvature
Alulis’ treatment [51] of the line of contact along a tooth depends upon the ap-
plicability of the models used for determining the Zero Error Cylinder Relationship and
the actual gear. Said relationship gives an accurate finite element size given an element
depth and a Hertzian deformation width for the forces along the line of contact. The
Hertzian deformation is a function of the magnitude of the applied load and the radius
of curvature.
A relation the radius of curvature across the line of contact must be determined.
The method used to determine a relationship for the radius of curvature across the line of
contact is a brute force geometric transformation. Since the spur profile is proportional
to the roll angle and base radius, a similar correlation is assumed. Points on and about
the line of contact are mapped to global cartesian coordinates and these coordinates are
analyzed.
Any three points in cartesian space are contained within a circle of a unique
radius. The solution to the radius of a circumcircle for a triangle of three points in a
plane is shown by Weisstein [60]. The solution of the circle is shown in Equation B.7.
¯ ¯
¯ ¯
¯ R2 x y 1 ¯
¯ ¯
¯ R12 x1 y1 1 ¯
¯ ¯
¯ ¯=0 (B.7)
¯
¯ R22 x2 y2 1 ¯
¯
¯ ¯
¯ R32 x3 y3 1 ¯
Where any Ri2 = x2i + yi2 . Weisstein shows how the radius and center can be
found by using the equation of a quadratic curve.
ax2 + cy 2 + dx + ey + f = 0 (B.8)
Where a = c. Note the absence of a cross term. The new center is found by
Equation B.9
[x0 , y0 ] = [−d/2a, −e/2a] (B.9)
The radius can be found by completing the square on Equation B.8 and solving.
s
d2 + e2 f
R= − (B.10)
4a2 a
102
The constants of the quadratic curve can be found by the Equations B.11.
¯ ¯
¯ ¯
¯x1 y1 1¯
¯ ¯
a = ¯¯x2 y2 1¯¯ (B.11a)
¯ ¯
¯x3 y3 1¯
¯ ¯
¯ 2 ¯
¯R1 y1 1¯
¯ 2 ¯
d = − ¯¯R y2 1¯¯ (B.11b)
¯ 22 ¯
¯R y3 1¯
¯ 3 ¯
¯ 2 ¯
R
¯ 1 x1 1¯
¯ 2 ¯
e = ¯R x2 1¯¯
¯ (B.11c)
¯ 22 ¯
¯R x3 1¯
¯3 ¯
¯ 2 ¯
¯R1 x1 y1 ¯
¯ ¯
f = − ¯¯R2 x2 y2 ¯¯ (B.11d)
¯ 22 ¯
¯R x3 y3 ¯
3
Solving the test gear with the above gives a relation that shows a strong linear
correlation between the radius of curvature and the roll angle (Figure B.1).
The proportionality for the specific geometry of the gear of analysis resulted in
Equation B.12.
ρc = ²bx (B.12)
Where bx = Rb · 0.955447.
103
Curvature Regression
30
28
26
Radius of Curvature (mm)
24
22
20
18
16
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4
Roll Angle (rad)
Fig. B.1. The solution of the radius analysis and the linear regression of the analysis
are nearly collinear.
104
Appendix C
The Analysis Primary Code
The work for the Fourier Null Matching Technique is performed in the code base
for the project. Visual Basic was chosen due to existing gear analysis code already written
in VB. It was determined that porting the existing code would be overly burdensome and
would serve to isolate gear analysis code libraries which was to be avoided if possible.
Primary functions and subroutines are shown. Thousands of ancillary lines of
code are not displayed here, but the purpose and function of the primary subroutines
should be apparent. All variables are global within the code library.
C.1 Main
The main subroutine establishes the gear geometry and sets up the optimization.
Sub MainLoop(g As GearGeom, ge As GearData, m As MatData)

Dim EPS#, Ha#, up#(), A#, u
’--------INPUT_VAR---------------------------------------------’
N = g.N
Phi = g.Phi
Psi = g.Psi
F = g.F ’
R = g.R
Rh = g.Rh ’hole radius
Ce = 2 * (1 - m.Nu ^ 2) / m.Elas
Nu = m.Nu
Elas = m.Elas
’----END-INPUT_VAR---------------------------------------------’
HoH = -g.rhh * 2 - 1 ’hand of helix

105
Psib = Atan(Tan(Psi) * Cos(Phi))

phin = Atan(Cos(Psi) * Tan(Phi))
thetap = Atan(Tan(Psi) * Sin(phin)) ’loadang
zang = Atan(Sin(Phi) * Tan(Psib)) ’ Angle of loading in z-plane
Rb = R * Cos(Phi)
P = N / 2 / R
pc = Pi / P
rr = R - 1.25 / P
Ra = R + 1 / P
phia = Acos(Rb / Ra)

ea = Tan(phia)
tha = ea - phia
eu = Ceil(ea / Pi * ths) + Ceil(pr / Pi * ths)
el = 2 * Tan(Phi) - ea ’ for a mating gear of the same size

phil = Atan(el)
betap = Pi / 2 / N
Ep = Tan(Phi)
psir = 2 * betap
pr = (Ep) - Phi ’ pr = theta = eps - phi : rotate through Theta
delta = 2 * Pi * Rb / N
Qa = F * Tan(Psib) / delta ’Qt = Qa * QtQa
Lnom = Qa * delta ’L for when Qa = Qt
L = Rb * (ea - el)
Qt = L / delta
Beta = Sin(Phi)
gamma = HoH * Beta * Tan(Psib)
D = Beta * L
betas = 32 / 63
ks = 0.25
106
alpha = 1
yb = Rb * Sin(Phi)
xb = Rb * Cos(Phi)
Call Str2ValMat(up, " 1,5,10,15,20,25")

u0 = 0.000001
Call Matsprd(up, u0)
h = 0.000025
Call InitFEAProg
Call GearMake(up)
End Sub
C.2 Gear Design Control
The subroutine GearMake controls the overall structure and the loop control for
the interative geometric solution.
Private Sub GearMake(u_in() As Double)
Call SetParam
Call InitFEA
For si = TNsL + 1 To TNsL * 2 - 1 ’si = TNsL To 1 Step -1 ’

For uI = 1 To UBound(u_in)
u = u_in(uI)
loadflag = u / u0 - 1
Call SetLocConst
Call OpenWIP00
If loadflag < 3 Then Call GraphLegL(True)
107
If InStr(1, ",4,9,14,19,24,", Chr(44) + CStr(loadflag) + Chr(44)) Then

Call OpenWIP
Call GraphLegL(True)
End If
Call DevElemPos
If InStr(1, ",2,4,9,14,19,24,", Chr(44) + CStr(loadflag) + Chr(44)) Then

Call FallOffAndReMap
Call GraphLegL(True)
End If
Call MakeFEModel
Call AllFEA
Call SolveCompliance
the = the + 1 ’iteration index
zl = Convergence(YP, ypL)
If zl < 0.95 And the < 150 Then
’not converged
SMatCopy ypL, YP
Call NotConverged
If loadflag >= 4 Then
Call FindTheta
Else
Call FindLocEnd
End If
Call FillKTCys
uI = uI - 1 ’rerun this loading
Else
’converged
If loadflag < 3 Then
Call FindLocEnd
Else
108
Call FindTheta
End If
Call FillKTCys
yBll = YBl
yBul = YBu
End If
SaveText LoadText, marker + "/" + hilite
Next uI
Next si
End Sub
C.3 Parameter Setup
The subroutine SetParam sets the basic bounds, the construction angles, and
positions for each line of contact. It also sets the base Modified Legendre loading profiles.
Private Sub SetParam()
su = D / 2 / Beta + F * Tan(Psib) * 0.5 ’ s at D/2 and -F/2

sl = -su ’-D / 2 / beta - F * Tan(psib) * 0.5
euu = su / Rb + Ep ’ roll angle at y=0 for s_upper
ell = sl / Rb + Ep
marker = CStr(N) + "_" + CStr(R) + "_" + CStr(F)
deltas = delta / 10 ’ stock is delta/10

TNsL = delta / deltas ’this assumes a stop of delta
nei = F / 10 ’nominal element width -- stock is 10

Nba = 4 ’the number of elements below and above the e and 2e width elements
109
thetaf = Asin(Rf / (Rf + rr))

thetaw = betap + pr + thetaf
Bx = rr * Sin(thetaw)
By = rr * Cos(thetaw)
OV = By - Bx
thetai = 0
du = 0.000005
Str2IntMat ElMap, "1,17,5, 12,16, 4,20,8" ’node number mapping

Ple = 75
Wei(0, 1) = -1 / 6: Wei(0, 3) = -1 / 6: Wei(0, 6) = -1 / 6
Wei(0, 8) = -1 / 6: Wei(0, 4) = 1 / 3: Wei(0, 5) = 1 / 3
Wei(0, 2) = 1 / 2: Wei(0, 7) = 1 / 2
Wei(1, 1) = -1 / 6: Wei(1, 3) = -1 / 6: Wei(1, 6) = -1 / 6

Wei(1, 8) = -1 / 6: Wei(1, 4) = 1 / 3: Wei(1, 5) = 1 / 3
Wei(1, 7) = 1 / 3: Wei(1, 2) = 2 / 3
Wei(2, 1) = -2 / 15: Wei(2, 3) = -2 / 15: Wei(2, 6) = -2 / 15

Wei(2, 8) = -2 / 15: Wei(2, 4) = 4 / 15: Wei(2, 5) = 4 / 15
Wei(2, 2) = 1 / 2: Wei(2, 7) = 1 / 2
Else
Wei(0, 1) = -1 / 6: Wei(0, 3) = -1 / 6: Wei(0, 6) = -1 / 6
Wei(0, 8) = -1 / 6: Wei(0, 4) = 1 / 3: Wei(0, 5) = 1 / 3
Wei(0, 2) = 1 / 2: Wei(0, 7) = 1 / 2 ’first leg loading
Wei(1, 1) = 0: Wei(1, 3) = 0: Wei(1, 6) = 0: Wei(1, 8) = 0

Wei(1, 4) = 0: Wei(1, 5) = 0
Wei(1, 2) = r3 / 6: Wei(1, 7) = -r3 / 6 ’ second loading only
Wei(2, 1) = r5 / 30: Wei(2, 3) = r5 / 30: Wei(2, 6) = r5 / 30

Wei(2, 8) = r5 / 30: Wei(2, 4) = -r5 / 15: Wei(2, 5) = -r5 / 15
Wei(2, 2) = 0: Wei(2, 7) = 0 ’third loading only
110
End If
End Sub
C.4 Initiate the Finite Element Programs
The subroutine InitFEAProg starts the finite element analysis programs.
Private Sub InitFEAProg()
ChDir "C:\Program Files (x86)\FEA\Z88\z88110"

z88h = SmartShell("C:\Program Files (x86)\FEA\Z88\z88110\z88h.exe", 1, _
"Z88 Cuthill-McKee Program Z88H")
z88i1 = SmartShell("C:\Program Files (x86)\FEA\Z88\z88110\z88i1.exe", 1, _
"Z88 Iteration Solver Part 1 Z88I1")
z88i2 = SmartShell("C:\Program Files (x86)\FEA\Z88\z88110\z88i2.exe", 1, _
"Z88 Iteration Solver Part 2 Z88I2")
ChDir "C:\Program Files (x86)\Microsoft Visual Studio\VB98"
End Sub
C.5 Initialize the Finite Element Analysis
The subroutine InitFEA sets the control bounds for the axial coordinate on the
tooth face.
Private Sub InitFEA()

noFEA:
yu = F / 2
yl = -F / 2
YBu = F / 4
YBl = -F / 4
111
Z88s = " "

ReDim yAB(1 To 2, 1 To TNsL)
End Sub
C.6 Set the Location Constants
The subroutine SetLocConst finishes computing the variables for the line of con-
tact. It also sets the finite element axial bounds and placement.
Private Sub SetLocConst()

totalconstraints = 0
totalelements = 0
totalnodes = 0
Z88I1g2 = ""
Z88I1g3 = ""
Z88I2g2 = ""
sn = -(TNsL - si) * deltas ’s for this line of loading

fPrgs.t2 = "Running s = " + CStr(sn) + " where u = " + CStr(u)
fPrgs.v2 = 1
hilite = "s_" + CStr(sn) + "_u_" + CStr(u) + "_file.dat"

hilite0 = hilite
marker0 = marker + "/" + "s_" + CStr(sn) + "_u_" + CStr(u0) + "_file.dat"
en = sn / Rb + Ep ’epsilon_n --- roll angle of l.o.c. at center of tooth (y=0)

phien = Atan(en)
Rn = Rb / Cos(phien)
thetan = en - phien
betan = (betap + (Ep - Phi)) - thetan
yd = -Beta * sn / 2 / gamma ’y on the diagonal
112
’ makes sure the y values fit within ’ the -D/2<=z<=D/2 range

If yl < (-L / 2 - sn) / Tan(Psib) Then yl = (-L / 2 - sn) / Tan(Psib)
If yu > (L / 2 - sn) / Tan(Psib) Then yu = (L / 2 - sn) / Tan(Psib)
’ Note:
’ -F/2 = the end of the tooth
’ yl = the edge of the possible LOC bounded by D and F
’ yBl = the edge of the analysis section
’ yBu = upper edge of the analysis
’ yu = the upper LOC edge
’ F/2 = edge of the tooth
’
’ ne = is the number of elements possible within F and D
’ nem is the elements needed to finish width below -F/2
’ nep is the elements needed to finish width above F/2
snu = sn + HoH * Tan(Psib) * yu ’z/{\ss} for yu on the current l.o.c s = sn

snl = sn + HoH * Tan(Psib) * yl ’z/{\ss} for yl on the current l.o.c s = sn
If snu < snl Then Error 380

If snu > su Then snu = su: Error 380
If snl < sl Then snl = sl: Error 380
enu = snu / Rb + Ep ’epsilon for yu on the current l.o.c s = sn

enl = snl / Rb + Ep ’epsilon for yl on the current l.o.c s = sn
nep = CInt((F / 2 - yu) / nei + 0.15)

If nep = 0 And Abs((F / 2 - yu)) > 1E-16 Then nep = 1
If nep Then deltaep = (F / 2 - yu) / nep
nem = CInt((yl + F / 2) / nei + 0.15)

If nem = 0 And Abs((yl + F / 2)) > 1E-16 Then nem = 1
If nem Then deltaem = (yl + F / 2) / nem
End Sub
113
C.7 Open Saved Progress File
The subroutine OpenWIP uses the saved data for a line of contact at a given loading
to continue the loading iteration or estimating the next sequence of element loading.
Private Sub OpenWIP()

Dim NB#, sz$
If (loadflag = 4) Then
If FileExist(marker + "\" + hilite) Then
Call OpenText(LoadText, marker + "\" + hilite)
Else
If FileExist(marker + "\" + Replace(hilite, CStr(u), CStr(u0))) Then
If yBll = 0 Then
Call OpenText(LoadText, marker + "\" + Replace(hilite, _
CStr(sn), CStr(sn + deltas)))
yBll = ExtractLine(LoadText, 7)
yBul = ExtractLine(LoadText, 8)
End If
Call OpenText(LoadText, marker + "\" + Replace(hilite, CStr(u), _
CStr(u0)))
loadflag = loadflag - 1
Else
Call OpenText(LoadText, marker + "\" + Replace(hilite, CStr(sn), _
CStr(sn + deltas)))
End If
End If
End If
If InStr(1, ",9,14,19,24,", Chr(44) + CStr(loadflag) + Chr(44)) Then
If FileExist(marker + "\" + hilite) Then
If yBll = 0 Then
Call OpenText(LoadText, marker + "\" + _
Replace(hilite, CStr(u), CStr(u - du)))
yBll = ExtractLine(LoadText, 7)
yBul = ExtractLine(LoadText, 8)
114
End If
Call OpenText(LoadText, marker + "\" + hilite)
Else
Call OpenText(LoadText, marker + "\" + Replace(hilite, CStr(u), _
CStr(u - du)))
loadflag = loadflag - 1
End If
End If
Dim wq#
NB = ExtractLine(LoadText, 4)
If NB <> 0 Then
locS = ExtractLine(LoadText, 1)
locS2 = ExtractLine(LoadText, 2)
wq = ExtractLine(LoadText, 3)
If CSng(sn) <> CSng(wq) Then If si > 0 Then Debug.Assert 0
ne = NB
sz = ExtractLine(LoadText, 5)
Str2ValMat LegL, sz
Str2ValMat Yloc, sz
If sz <> "" Then
YBl = sz
YBu = ExtractLine(LoadText, 8)
e = ExtractLine(LoadText, 9)
alphaOmega = ExtractLine(LoadText, 10)
If loadflag > 3 Then thetai = e
End If
If yBll = 0 Then
yBll = YBl
yBul = YBu
115
End If
End If
End Sub
C.8 Graph the Legendre Loading Coefficients
The subroutine GraphLegL plots the current Legendre loading profile.
Private Sub GraphLegL(Optional ShowGraph As Boolean)

ReDim xp(1 To 1000)
GAULEG YBl, YBu, 1000, XA, txe
Lg3L = ne * 3
ln1 = ","
bi = 1
mi = locS
I = 1
d8 = 0
If Yloc(1) = 0 Then
Do While Yloc(mi) = 0
mi = mi + 1
Loop
End If
For bji = 1 To 1000
Do While XA(bji) > Yloc(mi + 2) ’And mi < net - 2
mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)
If I <= 1 Then
I = 1
ElseIf I > ne Then
I = ne
mi = mi - 2
Exit Do
Else
116
d8 = 1
J = I - 1
End If
Loop
d4 = Yloc(mi + 2) - XA(bji) ’distance from yi to end of element

d5 = Yloc(mi + 2) - Yloc(mi) ’total length of element
D1 = -d4 / d5 * 2 + 1 ’eta on the -1 to 1 range
xp(bji) = LegL(I * 3 - 2) + LegL(I * 3 - 1) * (r3) * D1 + _

LegL(I * 3) * (r5) / 2 * (3 * D1 ^ 2 - 1)
Next bji
If loadflag = 3 Then Matsprd xp, 5: loadflag = 4

If InStr(1, ",8,13,18,23,", Chr(44) + CStr(loadflag) + Chr(44)) Then _
Matsprd xp, (u / (u - du)): loadflag = loadflag + 1
If ShowGraph Then
frm3DGraph.Show
Plot frm3DGraph, XA, xp, "y: Face Location (m)", _
"Normal Loading (N/m)", "Saved Data"
End If
End Sub
C.9 Develop the Element Positions
The subroutine DevElemPos finishes initiating the control variables for the FEA
meshing routine.
Private Sub DevElemPos()
Dim snbu#, snbl#, enbu#, enbl#, neold%

neold = ne
117
ne = CInt((YBu - YBl) / nei + 0.15) ’element number

If ne = 0 Then
If YBu <> YBl Then ne = 1 ’’’ bc(2) <> bc(1)
End If
deltae = (YBu - YBl) / ne ’ element width in the unmodified region
nep = CInt((F / 2 - YBu) / nei + 0.15)

If nep = 0 And Abs((F / 2 - YBu)) > 1E-16 Then nep = 1
If nep Then deltaep = (F / 2 - YBu) / nep
nem = CInt((YBl + F / 2) / nei + 0.15)

If nem = 0 And Abs((YBl + F / 2)) > 1E-16 Then nem = 1
If nem Then deltaem = (YBl + F / 2) / nem
net = 2 * (nep + nem + ne) + 1

’total number of transverse node groups as measured axially
locS = 1 + 2 * nem
locS2 = net - 2 * nep
Else
If YBu - bc(2) <> bc(1) - YBl Then
YBu = bc(2) + thetai
YBl = bc(1) - thetai
End If

Do While thetai > (bc(1) - yl) / 4
thetai = thetai / 10
Loop
End If
ne = CInt((YBu - YBl) / nei + 0.15) ’element number

If ne = 0 Then
If YBu <> YBl Then ne = 1
118
End If
deltae = (YBu - YBl) / ne ’ element width in the unmodified region
nep = CInt((F / 2 - YBu) / nei + 0.15)

If nep = 0 And Abs((F / 2 - YBu)) > 1E-16 Then nep = 1
If nep Then deltaep = (F / 2 - YBu) / nep
nem = CInt((YBl + F / 2) / nei + 0.15)

If nem = 0 And Abs((YBl + F / 2)) > 1E-16 Then nem = 1
If nem Then deltaem = (YBl + F / 2) / nem
’total number of transverse node groups as measured axially
net = 2 * (nep + nem + ne) + 1
locS = 1 + 2 * nem
locS2 = net - 2 * nep
End If
ReDim elInd(0 To net, 1 To 5)

Debug.Print "s = ", sn
Debug.Print "element dist", nem, ne, nep
SMatCopy Yloc2, Yloc

ReDim eLoc(1 To net)
ReDim Preserve Yloc(1 To net)
If HoH = 1 Then ’if g.rhh

eLoc(1) = enl
Else
eLoc(1) = enu
eii = -eii
End If
SMatCopy YP, Yloc
Yloc(1) = -F * 0.5
For mi = 2 To net
If mi <= (2 * nem + 1) Then
119
Yloc(mi) = Yloc(mi - 1) + deltaem / 2

ElseIf mi <= (2 * ne + 2 * nem + 1) Then
Yloc(mi) = Yloc(mi - 1) + deltae / 2
Else
Yloc(mi) = Yloc(mi - 1) + deltaep / 2
End If
eLoc(mi) = (sn + HoH * Tan(Psib) * Yloc(mi)) / Rb + Ep
If Abs(Yloc(mi)) < 1E-16 Then Yloc(mi) = 0#
Next mi
Debug.Assert Abs(Yloc(net) - F / 2) < 1E-16
End Sub
C.10 Map the Loading Profile on the Current Element Set
The subroutine FallOffAndRemap takes the current load profile and maps it to
the current finite element map.
Private Sub FallOffAndReMap()

Dim ml#, mu#
SMatCopy YP, Yloc
K = ne * 3
D1 = 1
ai = 2
a0 = 50
ReDim Yr(1 To K)
For yi = 1 To ne
b1 = Yloc(2 * (nem + yi) - 1)
B2 = Yloc(2 * (nem + yi) + 1)
ye = 0
GAULEG b1, B2, a0, tye, txe

ReDim Ya(1 To a0)
For mi = 1 To a0
120
d4 = tye(mi)
Do While XA(ai) < d4 And ai < 1000
ai = ai + 1
Loop
d12 = (xp(ai) - xp(ai - 1)) / (XA(ai) - XA(ai - 1)) ’m
d16 = xp(ai) - d12 * XA(ai) ’b
Ya(mi) = d12 * d4 + d16 ’ y = mx+b
Next mi
bji = 0: ye = 0
d20 = 1
For mi = 1 To a0
ye = ye + Ya(mi) * d20 * txe(mi)
’ sum(f(x)*Qn(x)*w)
’note: This is gausian quadrature for int(f Qn) over the element
Next mi
Yr(3 * yi - 2) = ye / (B2 - b1)
’to unscale the gaussian weights to -1>x>1; don’t need 0.5 *
bji = 1: ye = 0
d5 = B2 - b1 ’total length of element
For mi = 1 To a0
d4 = B2 - tye(mi) ’distance from a to end of element
d4 = -d4 / d5 * 2 + 1 ’eta on the -1 to 1 range
d20 = r3 * d4
Next mi
Yr(3 * yi - 1) = ye / (d5)
bji = 2: ye = 0
d5 = B2 - b1 ’total length of element
For mi = 1 To a0
d4 = B2 - tye(mi) ’distance from a to end of element
d4 = -d4 / d5 * 2 + 1 ’eta on the -1 to 1 range
d20 = 0.5 * r5 * (3 * d4 ^ 2 - 1)
Next mi
121
Yr(3 * yi) = ye / (d5)
Next yi
SMatCopy y, LegL
SMatCopy LegL, Yr
End Sub
C.11 Find the Hertzian Width
The subroutine Find_b uses the Zero Error Cylinder Relationship to find the
appropriate finite element size for a given position on the tooth and load magnitude.
Private Sub Find_b()

Dim thetai#
eii = eLoc(yi)
If eii >= ea + 0.000001 Then
ei = ea - 0.01
ElseIf eii < el Then
If ei < el Or ei > el + deltae Then ei = el + 0.01
’else keep ei the same
Else
ei = eii
If 0 = locS And yi > 2 * nem Then locS = yi
’locs is the planar element number beginning the analysis range
End If
If yi > 2 * nem And yi <= net - 2 * nep Then locN = locN + 1
phii = Atan(ei)
thetai = ei - phii
betai = betap + (Ep - Phi) - thetai
Ri = Rb / Cos(phii)
Kd = Rb * ei * 0.94475 ’ ROC = Rb * ei * bx’

122
If loadflag = 0 Then
’ "PL" = p = P/L
PL = ZeroLoad(Yloc(yi))
ElseIf loadflag = 1 Then ’lf=1 picking up where we left off
If locS And yi <= net - 2 * nep Then
ReDim LegL(1 To 3 * ne)
For I = 1 To ne
LegL(I * 3 - 2) = KTCys(si, I, 2) * u
LegL(I * 3 - 1) = KTCys(si, I, 3) * u
LegL(I * 3) = KTCys(si, I, 4) * u
Next I
PL = SavedLoad(yi, locS, Yloc, LegL)
’note that a loading profile has to be N/m
Else
PL = 0
End If
loadflag = 3
Debug.Assert 0
ElseIf loadflag >= 2 Then
If locS And yi <= net - 2 * nep Then
PL = SavedLoad(yi, locS, Yloc, LegL)
Else
PL = 0
End If
End If
b = 1.6 * Sqrt(PL * Kd * Ce) ’ Hertzian Formula

If b < 0.000001 Then b = 0.000001 ’ enforce lower bound ’
End Sub
123
C.12 Mesh the Finite Element Model
The subroutine MakeFEModel creates the finite element mesh for the current gear
tooth iteration.
Private Sub MakeFEModel()
ReDim eic(1 To 3, 1 To net)

ReDim ebc(1 To net, 1 To 3)
locS = 0
locN = 0
End If
NX = 8 ’ NX is the number of elements across the tooth (tip)

cg = Rb
For yi = 1 To net
Call Find_b
If (Yloc(yi) >= YBl And Yloc(yi) <= YBu) Or _
(YBu - YBl < deltae / 2 And YBu - Yloc(yi) < deltae / 2 And _
YBu - Yloc(yi) > 0) Then
NX = Ceil(7 / 20 * Ri * Sin(betai) / b)
’ = ti/c_max = required number of elements across the tooth
Nxg = Max0(Nxg, NX)
c = Dmin1(0.25 * Ri * Sin(betai), 40 / 7 * b)
’ min of ti/8 and b requirement
cg = Dmin1(cg, c)
eic(1, yi) = eLoc(yi)

eic(2, yi) = c
eic(3, yi) = NX
End If
Next yi
124
locS = 0
locN = 0
NX = Ceil(8 - 6 / Nxg) + 6
If odd(NX) Then NX = NX + 1
N2D = 11 * (13 + NX) + (Nba * 2 + 9) * (3 * NX + 2)

’ number of nodes in a 2d transverse plane
ReDim inner(1 To 2, 1 To 13 + NX)
ReDim outer(1 To 2, 1 To 13 + NX)
Nxg2 = Nxg / 2 ’run the proportionality on the halfside of the tooth
J = 1
For angle = 90 + 40 To -190 Step -40
If angle = 130 Or angle = 90 Then
’the starting point for the circles (at node 1)
I = J
LinSpace arc, angle, angle - 40, 7 + NX
For ai = 1 To NX + 5 Step 2
inner(1, J) = arc(ai) * Pi / 180 ’element corner nodes
inner(2, J) = arc(ai + 1) * Pi / 180
’interstitial nodes following the cornernode_j clockwise
J = J + 1
Next ai
TW = thetaw / Pi * 180
If angle = 130 Then
LinSpace arc, angle, angle - 40 + TW, 7
For ai = 1 To 5 Step 2
outer(1, I) = arc(ai) * Pi / 180
outer(2, I) = arc(ai + 1) * Pi / 180
I = I + 1
Next ai
LinSpace arc, -Bx, 0, NX + 1 ’ x coord of wedge nodes
For ai = 1 To NX - 1 Step 2
125
outer(1, I) = arc(ai)
outer(2, I) = arc(ai + 1)
I = I + 1
Next ai
Else
LinSpace arc, OV, By, NX + 1 ’ y coord of wedge nodes
For ai = 1 To NX - 1 Step 2
outer(1, I) = arc(ai)
outer(2, I) = arc(ai + 1)
I = I + 1
Next ai
LinSpace arc, angle - TW, angle - 40, 7

For ai = 1 To 5 Step 2
outer(1, I) = arc(ai) * Pi / 180
outer(2, I) = arc(ai + 1) * Pi / 180
I = I + 1
Next ai
End If
Else
inner(1, J) = angle * Pi / 180
inner(2, J) = (angle - 20) * Pi / 180
outer(1, J) = angle * Pi / 180
outer(2, J) = (angle - 20) * Pi / 180
J = J + 1
End If
Next angle
’ Now find the radius points

gel = J - 1 ’number of gear elements around the gear
angA = Pi - betap - pr
126
angB = angA + Pi / 2 - thetaf
Cloc(1) = (rr + Rf) * Sin(thetaw) ’X fillet center location

Cloc(2) = (rr + Rf) * Cos(thetaw) ’Y fillet center location
LinSpace arc, angA, angB, 7
For J = 1 To 3
I = (J - 1) * 2 + 1
RadLoc(1, J, 1) = Rf * Cos(arc(I)) + Cloc(1) ’corner node X

RadLoc(1, J, 2) = Rf * Sin(arc(I)) + Cloc(2) ’corner node Y
RadLoc(2, J, 1) = Rf * Cos(arc(I + 1)) + Cloc(1) ’interstitial nodes X

RadLoc(2, J, 2) = Rf * Sin(arc(I + 1)) + Cloc(2) ’interstitial nodes Y
Next J
’order up the node matrix part for the gear

Dim gnodes2D#()
ReDim gnodes2D(1 To 2, 1 To (11 * (13 + NX) + 8 + NX * 12))
For J = 1 To gel
iL = (J - 1) * 11 + 1 ’node (gnode) index for first position
If Abs(outer(1, J)) < 0.01 Then
LinSpace tx, (outer(1, J)), Rh * Cos(inner(1, J)), 7
If Abs(outer(1, NX + 8 - J)) > 0.1 Then
LinSpace ty, rr * Sin(outer(1, NX + 8 - J)), Rh * _
Sin(inner(1, J)), 7
Else
LinSpace ty, (outer(1, NX + 8 - J)), Rh * Sin(inner(1, J)), 7
End If
ElseIf Abs(outer(1, J)) < 0.1 Then
LinSpace ty, (outer(1, J)), Rh * Sin(inner(1, J)), 7
If outer(1, NX + 8 - J) > 0.01 Then
LinSpace tx, 0, 0, 7
Else
LinSpace tx, -(outer(1, NX + 8 - J)), Rh * Cos(inner(1, J)), 7
127
End If
Else
LinSpace tx, rr * Cos(outer(1, J)), Rh * Cos(inner(1, J)), 7
LinSpace ty, rr * Sin(outer(1, J)), Rh * Sin(inner(1, J)), 7
End If
For I = 0 To 6
gnodes2D(1, I + iL) = tx(I + 1)
gnodes2D(2, I + iL) = ty(I + 1)
Next I
iL = (J - 1) * 11 + 8
If Abs(outer(2, J)) < 0.01 Then
LinSpace tx, (outer(2, J)), Rh * Cos(inner(2, J)), 4
LinSpace ty, (outer(2, NX + 7 - J)), Rh * Sin(inner(2, J)), 4
ElseIf Abs(outer(2, J)) < 0.1 Then
LinSpace ty, (outer(2, J)), Rh * Sin(inner(2, J)), 4
LinSpace tx, -(outer(2, NX + 7 - J)), Rh * Cos(inner(2, J)), 4
Else
LinSpace tx, rr * Cos(outer(2, J)), Rh * Cos(inner(2, J)), 4
LinSpace ty, rr * Sin(outer(2, J)), Rh * Sin(inner(2, J)), 4
End If
For I = 0 To 3
Next I
Next J
Dim eProp#(), ePropI#()

LinSpace ePropI, 0, 1, NX + 1 ’element width proportionality
LinSpace eProp, 3, Nxg2, NX - 5 ’element width proportionality
ReDim Preserve eProp(1 To NX + 1)
For J = NX + 1 To 7 Step -1
eProp(J) = eProp(J - 6)
Next J
For J = 6 To 1 Step -1
128
eProp(J) = eProp(J + 1) - 1 / 2
Next J
VecsPrd eProp, 1 / Nxg2
LinSpace ty2, OV, RadLoc(1, 1, 2), 7

For J = 1 To 3
iL = (J - 1) * (NX * 3 + 2) + (11 * (13 + NX) + 1)
MidSpace tx, ePropI, eProp, (2 * J - 1) / 8, -RadLoc(2, 4 - J, 1), 0

MidSpace ty, ePropI, eProp, (2 * J - 1) / 8, RadLoc(2, 4 - J, 2), ty2(2 * J)
For I = 0 To NX / 2
gnodes2D(1, I + iL) = tx(2 * I + 1)
gnodes2D(2, I + iL) = ty(2 * I + 1)
Next I
For I = (NX / 2 + 1) To NX
gnodes2D(1, I + iL) = -tx(2 * (NX - I) + 1)
gnodes2D(2, I + iL) = ty(2 * (NX - I) + 1)
Next I
iL = (J - 1) * (NX * 3 + 2) + (11 * (13 + NX) + 2 + NX)
MidSpace tx, ePropI, eProp, J / 4, -RadLoc(1, 4 - J, 1), 0

MidSpace ty, ePropI, eProp, J / 4, RadLoc(1, 4 - J, 2), ty2(2 * J + 1)
For I = 0 To NX
Next I
For I = NX + 1 To 2 * NX
gnodes2D(1, I + iL) = -tx(2 * NX + 1 - I)
gnodes2D(2, I + iL) = ty(2 * NX + 1 - I)
Next I
Next J
129
ReDim Preserve eProp(1 To NX * 2 + 1)

VecsPrd eProp, 0.5
For I = NX + 2 To NX * 2 + 1
eProp(I) = 1 - eProp(-I + 2 * NX + 2)
Next I
LinSpace ePropI, 0, 1, 2 * NX + 1
LinSpace arc, Rb * Sin(betap + pr), RadLoc(1, 1, 1), 3

LinSpace ty2, Rb * Cos(betap + pr), RadLoc(1, 1, 2), 3
iL = (J - 1) * (NX * 3 + 2) + (11 * (13 + NX) + 1)
MidSpace tx, ePropI, eProp, ((2 * J - 1) / 8), -arc(2), arc(2)

For I = 0 To NX
gnodes2D(1, I + iL) = tx(2 * I + 1)
gnodes2D(2, I + iL) = ty2(2)
Next I
iL = (J - 1) * (NX * 3 + 2) + (11 * (13 + NX) + 2 + NX)

RangeSpace tx, eProp, -arc(1), arc(1)
For I = 0 To 2 * NX
gnodes2D(2, I + iL) = ty2(1)
Next I
For yi = 1 To net
fPrgs.t1.Text = "Building node set " + CStr(yi) + " out of " + CStr(net)
Call Find_b
c = 2 * Ri * Sin(betai) / NX
ebcCalc e, b, c
ebc(yi, 1) = e: ebc(yi, 2) = b: ebc(yi, 3) = c
e = e / Cos(thetap)
’ebc gives e value perp to line of loading -- adjust to xz plane
130
’MMMMMMMMMMMMMMMMMMM Create the mesh on the tooth MMMMMMMMMMMMMM

ReDim invLoc(1 To 6 + 2 * Nba)
If (ei ^ 2) < (7 * e / Rb) Then
e = ei ^ 2 * Rb / 10
End If
inveu = Sqrt(ei ^ 2 + 7 * e / Rb)

invel = Sqrt(ei ^ 2 - 7 * e / Rb)
VecFill epsi, 0, invel ^ 2, -Nba - 1

For mi = 1 To Nba + 1
invLoc(mi) = Sqrt(epsi(mi))
Next mi
invLoc(Nba + 2) = Sqrt(ei ^ 2 - 3 * e / Rb)

invLoc(Nba + 3) = Sqrt(ei ^ 2 - e / Rb)
invLoc(Nba + 4) = Sqrt(ei ^ 2 + e / Rb)
invLoc(Nba + 5) = Sqrt(ei ^ 2 + 3 * e / Rb)
VecFill epsi, inveu ^ 2, ea ^ 2, -(Nba + 1)

For mi = 1 To Nba + 1
invLoc(Nba + 5 + mi) = Sqrt(epsi(mi))
Next mi
If odd(yi) Then ’no need to do the interstitial nodes for this location
ReDim intstitLoc(1 To 2 * Nba + 5)
VecFill epsi, 0, invel ^ 2, -(2 * Nba + 1)

For mi = 1 To Nba
intstitLoc(mi) = Sqrt(epsi(2 * mi))
Next mi
intstitLoc(Nba + 1) = Sqrt(ei ^ 2 - 5 * e / Rb)

intstitLoc(Nba + 2) = Sqrt(ei ^ 2 - 2 * e / Rb)
131
intstitLoc(Nba + 3) = ei
intstitLoc(Nba + 4) = Sqrt(ei ^ 2 + 2 * e / Rb)
intstitLoc(Nba + 5) = Sqrt(ei ^ 2 + 5 * e / Rb)
VecFill epsi, inveu ^ 2, ea ^ 2, -(2 * Nba + 1)

For mi = 1 To Nba
intstitLoc(Nba + 5 + mi) = Sqrt(epsi(2 * mi))
Next mi
End If
’order up the node matrix part for the tooth

SMatCopy tnodes2D, gnodes2D
ReDim Preserve tnodes2D(1 To 2, 1 To N2D)
iu = UBound(intstitLoc)
For J = 1 To iu
iL = (J + 3) * (NX * 3 + 2) + (11 * (13 + NX) + 1)
ModInv xn, yn, intstitLoc(J), betap, Rb, pr

RangeSpace tx, eProp, -xn, xn
For I = 0 To NX
tnodes2D(1, I + iL) = tx(2 * I + 1)
tnodes2D(2, I + iL) = yn
Next I
iL = (J + 3) * (NX * 3 + 2) + (11 * (13 + NX) + 1) + (NX + 1)

ModInv xn, yn, invLoc(J + 1), betap, Rb, pr
RangeSpace tx, eProp, -xn, xn ’, 17
For I = 0 To 2 * NX ’16
tnodes2D(1, I + iL) = tx(I + 1)
tnodes2D(2, I + iL) = yn
Next I
Next J
’first, rotate the points to the correct postion

132
NodesRot tnodes2D, (betap + Ep - eii)
’now save mat and index to make the next 2d mats and make the elements
If yi = 1 Then
lN = 0
SMatCopy lastnodes, tnodes2D
’ add last nodes to the element text file
J = N2D
For I = 1 To J
totalnodes = totalnodes + 1
Z88I1g2 = Z88I1g2 + Z88s + NStr(totalnodes, 5) + Z88s + "3" + _
Z88s + NStr(lastnodes(1, I), 21) + Z88s + NStr(Yloc(yi), 21) + _
Z88s + NStr(lastnodes(2, I), 21) + vbCrLf
Next I
For I = 1 To 13 + NX
xi = (I - 1) * 11
totalconstraints = totalconstraints + 6
Z88I2g2 = Z88I2g2 + Z88s + NStr(xi + 7, 5) + Z88s + "1" + _
Z88s + "2" + Z88s + "0 " + vbCrLf
Z88s + "2" + Z88s + "0 " + vbCrLf
Z88I2g2 = Z88I2g2 + Z88s + NStr(xi + 7, 5) + Z88s + "3" _
+ Z88s + "2" + Z88s + "0 " + vbCrLf
+ Z88s + "2" + Z88s + "0 " + vbCrLf
+ Z88s + "2" + Z88s + "0 " + vbCrLf
+ Z88s + "2" + Z88s + "0 " + vbCrLf
Next I
I = N2D - (3 * NX + 2) * (Nba + 2) - (2 * NX + 1)
’2d node number at load line
elInd(0, 1) = I + (2 * NX + 1)
elInd(0, 2) = I
133
elInd(0, 3) = I - (NX + 1)
ElseIf even(yi) Then ’do the interstitial nodes

SMatCopy inttnodes, tnodes2D
ReDim MatInt(1 To N2D)
J = N2D
For I = 1 To (13 + NX)
For Ii = 1 To 7 Step 2
xi = (I - 1) * 11 + Ii
Z88s + NStr(inttnodes(1, xi), 21) + Z88s + NStr(Yloc(yi), 21) + _
Z88s + NStr(inttnodes(2, xi), 21) + vbCrLf
MatInt(xi) = totalnodes ’global node number
Next Ii
Next I
For I = (NX) + (11 * (13 + NX) + 1) To N2D - 5 Step (3 * NX + 2) ’26
For Ii = 1 To (2 * NX + 1) Step 2 ’17
xi = I + Ii
Z88s + NStr(inttnodes(1, xi), 21) + Z88s + NStr(Yloc(yi), 21) + _
Z88s + NStr(inttnodes(2, xi), 21) + vbCrLf
MatInt(xi) = totalnodes
Next Ii
Next I
For I = 1 To (13 + NX) ’21

xi = (I - 1) * 11 ’ 7 only
Z88I2g2 = Z88I2g2 + Z88s + NStr(MatInt(xi + 7), 5) + Z88s + "1" _
+ Z88s + "2" + Z88s + "0 " + vbCrLf
+ Z88s + "2" + Z88s + "0 " + vbCrLf
134

+ Z88s + "2" + Z88s + "0 " + vbCrLf
Next I
Else
’ add last nodes to the element text file
’node labelling here
TN = totalnodes
For I = 1 To N2D ’j
Z88s + NStr(tnodes2D(1, I), 21) + Z88s + NStr(Yloc(yi), 21) + _
Z88s + NStr(tnodes2D(2, I), 21) + vbCrLf
Next I
Dim g3L1$, g3L2$

For J = 1 To (NX + 13)
For I = 1 To 3
totalelements = totalelements + 1
’matrix of 2d element numbers
eleMat(1) = 11 * J + I * 2 - 1 ’12T
eleMat(2) = 11 * (J - 1) + I * 2 - 1 ’1T
eleMat(3) = 11 * (J - 1) + I * 2 - 1 ’1L
eleMat(4) = 11 * J + I * 2 - 1 ’12L
eleMat(5) = 11 * J + I * 2 + 1 ’14T
eleMat(6) = 11 * (J - 1) + I * 2 + 1 ’3T
eleMat(7) = 11 * (J - 1) + I * 2 + 1 ’3L
eleMat(8) = 11 * J + I * 2 + 1 ’14L
eleMat(9) = 11 * (J - 1) + I + 7 ’8T
eleMat(10) = 11 * (J - 1) + I * 2 - 1 ’1I
eleMat(11) = 11 * (J - 1) + I + 7 ’8L
eleMat(12) = 11 * J + I * 2 - 1 ’12I
eleMat(13) = 11 * (J - 1) + I + 8 ’9T
eleMat(14) = 11 * (J - 1) + I * 2 + 1 ’3I
eleMat(15) = 11 * (J - 1) + I + 8 ’9L
eleMat(16) = 11 * J + I * 2 + 1 ’14I
135
eleMat(17) = 11 * J + I * 2 ’13T
eleMat(18) = 11 * (J - 1) + I * 2 ’2T
eleMat(19) = 11 * (J - 1) + I * 2 ’2L
eleMat(20) = 11 * J + I * 2 ’13L
If J = (NX + 13) Then ’21

For Ii = 1 To 20
If eleMat(Ii) > (11 * (13 + NX)) Then _
eleMat(Ii) = eleMat(Ii) - (11 * (13 + NX))
Next Ii
End If
g3L1 = Z88s + NStr(totalelements, 4) + Z88s + "10" + Z88s _

+ "node no. " + NStr(totalelements, 4)
g3L2 = Z88s + NStr(eleMat(1) + TN, 5) + Z88s _

+ NStr(eleMat(2) + TN, 5) + Z88s + NStr(eleMat(3) + lN, 5) _
+ Z88s + NStr(eleMat(4) + lN, 5)
g3L2 = g3L2 + Z88s + NStr(eleMat(5) + TN, 5) + Z88s + _
NStr(eleMat(6) + TN, 5) + Z88s + _
NStr(eleMat(7) + lN, 5) + Z88s + NStr(eleMat(8) + lN, 5)
NStr(MatInt(eleMat(10)), 5) + Z88s + _
NStr(eleMat(11) + lN,5) + Z88s + NStr(MatInt(eleMat(12)),5)
NStr(eleMat(18) + TN, 5) + Z88s + _
NStr(eleMat(19) + lN, 5) + Z88s + NStr(eleMat(20) + lN, 5)
Lg3L = Len(g3L1 + vbCrLf + g3L2 + vbCrLf)

Z88I1g3 = Z88I1g3 + g3L1 + vbCrLf + g3L2 + vbCrLf
Next I
Next J
136
For J = (NX) + (11 * (13 + NX) + 1) To N2D - 5 Step (3 * NX + 2)

For I = 1 To (2 * NX - 1) Step 2
xi = I + J
totalelements = totalelements + 1
’matrix of 2d element numbers
eleMat(1) = xi + 2
eleMat(2) = xi
eleMat(3) = xi
eleMat(4) = xi + 2
eleMat(5) = xi - (3 * NX)
eleMat(6) = xi - (3 * NX + 2)
eleMat(7) = xi - (3 * NX + 2)
eleMat(8) = xi - (3 * NX)
eleMat(9) = xi + 1
eleMat(10) = xi
eleMat(11) = xi + 1
eleMat(12) = xi + 2
eleMat(13) = xi - (3 * NX + 1)
eleMat(14) = xi - (3 * NX + 2)
eleMat(15) = xi - (3 * NX + 1)
eleMat(16) = xi - (3 * NX)
eleMat(17) = (J - (2 * NX - 1) + xi) * 0.5
eleMat(18) = (J - (2 * NX + 1) + xi) * 0.5
eleMat(19) = (J - (2 * NX + 1) + xi) * 0.5
eleMat(20) = (J - (2 * NX - 1) + xi) * 0.5
If J = (NX) + (11 * (13 + NX) + 1) Then

b1 = (11 * (13 + NX) + 1) - (2 * NX + 1) - 2
B2 = (11 * (13 + NX) + 1) - (2 * NX + 1) - 3
For Ii = 1 To 20
If eleMat(Ii) < (11 * (13 + NX) + 1) Then
If even(eleMat(Ii)) Then
eleMat(Ii) = (eleMat(Ii) - B2) / 2 * 11 + 19
Else
137
eleMat(Ii) = (eleMat(Ii) - b1) / 2 * 11 + 23

End If
End If
Next Ii
End If
g3L1 = Z88s + NStr(totalelements, 4) + Z88s + "10" + _

Z88s + "node no. " + NStr(totalelements, 4)
g3L2 = Z88s + NStr(eleMat(1) + TN, 5) + Z88s + _

NStr(eleMat(2) + TN, 5) + Z88s + NStr(eleMat(3) + lN,5) + _
Z88s + NStr(eleMat(4) + lN, 5)
NStr(eleMat(6) + TN, 5) + Z88s + NStr(eleMat(7) + lN,5) + _
NStr(eleMat(18) + TN,5) + Z88s + NStr(eleMat(19) + lN,5) + _
Lg3L = Len(g3L1 + vbCrLf + g3L2 + vbCrLf)

Z88I1g3 = Z88I1g3 + g3L1 + vbCrLf + g3L2 + vbCrLf
Next I
Next J
For I = 1 To 13 + NX
xi = (I - 1) * 11 + TN
138
Z88s + "2" + Z88s + "0 " + vbCrLf

Z88s + "2" + Z88s + "0 " + vbCrLf
Z88s + "2" + Z88s + "0 " + vbCrLf
Z88s + "2" + Z88s + "0 " + vbCrLf
Z88s + "2" + Z88s + "0 " + vbCrLf
Z88s + "2" + Z88s + "0 " + vbCrLf
Next I
xi = (yi - 1) / 2
I = N2D - (3 * NX + 2) * (Nba + 2) - (2 * NX + 1)
’2d node number at load line
elInd(xi, 1) = I + (2 * NX + 1) + TN
elInd(xi, 2) = I + TN
elInd(xi, 3) = I - (NX + 1) + TN
elInd(xi, 4) = MatInt(I + (2 * NX + 1))
elInd(xi, 5) = MatInt(I - (NX + 1))
SMatCopy lastnodes, tnodes2D

lN = TN
End If
Next yi
locS2 = locS + locN - 1
’locs2 is the planar element number ending the analysis range
Debug.Assert (locN - 1) / 2 = ne
Z88I1g1 = Z88s + "3" + Z88s + CStr(totalnodes) + Z88s + _

CStr(totalelements) + Z88s + CStr(3 * totalnodes) + Z88s + "1" + _
Z88s + "0" + Z88s + "0" + Z88s + "0" + vbCrLf
Z88I1g4 = Z88s + "1" + Z88s + CStr(totalelements) + Z88s + CStr(Elas) + _
Z88s + CStr(Nu) + Z88s + "3" + Z88s + "0" + vbCrLf
139
El2D = totalelements / net

Z88I1g2 = Z88I1g1 + Z88I1g2 + Z88I1g3 + Z88I1g4
fnum2 = FreeFile
Open "z88i1.txt" For Output As #fnum2
Print #fnum2, Z88I1g2
Close #fnum2
End Sub
C.13 Perform the Finite Element Analysis
The subroutine AllFEA runs the top level finite element analysis.
Private Sub AllFEA()
If HoH = 1 Then I = 1 Else I = -1
UV(1) = -Cos(Psi) * Cos(phin)

UV(2) = I * Sin(Psi) * Cos(phin)
UV(3) = -Sin(phin) ’forces unit vector
Lg3L = CInt((locN - 1) / 2 - 0.1) ’’= the number of elements fully in range

ReDim comp(1 To Lg3L * 3, 1 To Lg3L * 3)
ReDim LegD(1 To Lg3L * 3)
Z88I2g2a = Z88I2g2
locS2 = locS + locN - 1

For bi = 1 To Lg3L
Call ForceFEModel
Call PrePlotFEA
Next bi
End Sub
140
C.14 Apply Forces to the Finite Element Model
The subroutine ForceFEModel creates the nodal loading parameters for the cur-
rent load profile.
Private Sub ForceFEModel()

yi = CInt(locS / 2.1) + bi
bQ(0) = 26 / deltae
bQ(1) = 26 / (r3) / deltae
bQ(2) = 26 / (r5) / deltae
Else
If bi <= ne Then
bQ(0) = LegL(3 * bi - 2)
bQ(1) = LegL(3 * bi - 1)
bQ(2) = LegL(3 * bi)
Else
bQ(0) = LegL(ne * 3 - 2)
bQ(1) = LegL(ne * 3 - 1)
bQ(2) = LegL(ne * 3)
End If
End If
’the numbers for the unit displacement are the numbers solved for in
’the overdetermined system.
For bji = 0 To 2
Z88I2g2 = Z88I2g2a
mi = InStr(1, Z88I2g2, vbCrLf) + 1
mi = InStr(mi, Z88I2g2, vbCrLf) - mi + 1
’fill the constraint lines with the loading data

For J = 1 To 8
141
If J < 6 Then
Ii = J
xi = yi
Else
Ii = J - 5
xi = yi - 1
End If
Z88I2g3 = ""
For I = 1 To 3
MB = Wei(bji, J) * Ple
Else
If bji = 0 Then
MB = Wei(bji, J) * bQ(bji)
Else
MB = Wei(0, J) * bQ(0) + Wei(bji, J) * bQ(bji)
End If
End If
Z88I2g3 = Z88I2g3 + Z88s + NStr(elInd(xi, Ii), 5) + Z88s + _
CStr(I) + Z88s + "1" + Z88s + NStr(UV(I) * MB, 24) + vbCrLf
Next I
ts = elInd(xi, Ii)
lN = 1: iu = Len(Z88s): xi = ts + 1
lN = Len(Z88I2g2) + 1
Ii = lN
Do While xi > ts
Ii = Ii - mi
Z88I2g1 = Mid(Z88I2g2, Ii, mi)
xi = Mid(Z88I2g1, 1, iu + 5)
Loop
If Ii < lN - mi Then
Z88I2g2 = Left(Z88I2g2, Ii + mi - 1) + Z88I2g3 + _
Mid(Z88I2g2, Ii + mi)
Else
Z88I2g2 = Z88I2g2 + Z88I2g3
142
End If
Next J
Dim fdt As Date, fmt As Date

Z88I2g1 = Z88s + CStr(totalconstraints + 24) + vbCrLf
Z88I2g2 = Z88I2g1 + Z88I2g2
fnum2 = FreeFile
Open "z88i2.txt" For Output As #fnum2
Print #fnum2, Z88I2g2
Close #fnum2 ’
Call RunFEA
Call ReadFEA
Next bji
End Sub
C.15 Execute the Finite Element Analysis
The subroutine RunFEA sends the data to the finite element programs for execu-
tion.
Private Sub RunFEA()

On Error Resume Next
FileCopy "z88i1.txt", "C:\Program Files (x86)\FEA\Z88\z88110\z88i1.txt"

FileCopy "z88.dyn", "C:\Program Files (x86)\FEA\Z88\z88110\z88.dyn"
I = SmartAppActivate("Z88 Cuthill-McKee Program Z88H")

I = I + SmartAppActivate("Z88 Cuthill-McKee Program Z88H")
’run Cuthill-McKee twice
I = I + SmartAppActivate("Z88 Iteration Solver Part 1 Z88I1")

143
I = I + SmartAppActivate("Z88 Iteration Solver Part 2 Z88I2")
FileCopy "C:\Program Files (x86)\FEA\Z88\z88110\z88o2.txt", "z88o2.txt"

FileCopy "C:\Program Files (x86)\FEA\Z88\z88110\z88i1.txt", "z88i1h.txt"
End Sub
C.16 Read the Finite Element Data
The subroutine ReadFEA takes the finite element solution and determines the
resultant displacement coefficients.
Private Sub ReadFEA()
fnum2 = FreeFile
Open "z88i1h.txt" For Input As #fnum2
fnum = FreeFile
Open "z88i1.txt" For Input As #fnum
’map the node numbers
NdNum = totalnodes
ElNum = totalelements
ReDim Nodes(1 To NdNum)
For I = 1 To NdNum + 1
Line Input #fnum, ln1
Line Input #fnum2, ln2
Next I
For I = 1 To ElNum

144
Str2IntMat CamI, ln1, 20

Str2IntMat CutI, ln2, 20
For K = 1 To 20
Nodes(CamI(K)) = CutI(K)
Next K
Next I
Close #fnum
Close #fnum2
fnum2 = FreeFile
Open "z88o2.txt" For Input As #fnum2
Line Input #fnum2, NewLine
ReDim arc(1 To 8)
ai = 0
ReDim NodeStr(1 To totalnodes)
For xi = 1 To totalnodes
NodeStr(xi) = Mid(NewLine, 7)
Next xi
Close #fnum2
For mi = 1 To Lg3L
iu = CInt(locS / 2.1) + mi
For xi = 8 To 1 Step -1
If xi < 6 Then I = iu: J = xi Else I = iu - 1: J = xi - 5
Str2ValMat ty2, NodeStr(Nodes(elInd(I, J)))
arc(xi) = UV(1) * ty2(1) + UV(2) * ty2(2) + UV(3) * ty2(3)
’dot product (the projection of the displacement onto the normal, UV)
145
Next xi
D1 = arc(1)
d17 = arc(2)
d5 = arc(3)
d12 = arc(4)
d16 = arc(5)
d4 = arc(6)
d20 = arc(7)
d8 = arc(8)
b1 = 0.25 * (D1 + d4 + d8 + d5) - 0.5 * (d12 + d16)

B2 = 0.5 * (d17 - d20)
b3 = -0.25 * (D1 + d4 + d8 + d5) + 0.5 * (d12 + d16 + d20 + d17)
a0 = b1 / 3 + b3
A1 = B2 / (r3)
A2 = 2 / 3 / (r5) * b1
’now put these in the Cmatrix

I = (mi - 1) * 3
J = bji + (bi * 3) - 2
If bji = 0 Then
LegD(I + 1) = a0
LegD(I + 2) = A1
LegD(I + 3) = A2
Else
a0 = a0 - LegD(I + 1)
A1 = A1 - LegD(I + 2)
A2 = A2 - LegD(I + 3)
End If
comp(I + 1, J) = a0 / bQ(bji)
146
comp(I + 2, J) = A1 / bQ(bji)
comp(I + 3, J) = A2 / bQ(bji)
Next mi
End Sub
C.17 Prepare the Finite Element Data for Plotting
The subroutine PrePlotFEA sets the displacement data from the finite element
analysis for graphical output.
Private Sub PrePlotFEA()
GAULEG yl, yu, 250, tye, txe

ReDim YP(1 To 250)
ln1 = ",": Size2d xp, ln1, True
If InStr(1, ln1, CStr(Lg3L) + ",") Then
ReDim Preserve xp(1 To Lg3L, 1 To 250)
Else
ReDim xp(1 To Lg3L, 1 To 250)
End If
mi = -1
For bji = 1 To 250
Do While tye(bji) > Yloc(mi + 2) ’And mi < net - 2
mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)

If I <= 0 Then I = 1
If I > ne Then I = ne: mi = mi - 2: Exit Do
Loop
d4 = Yloc(mi + 2) - tye(bji) ’distance from yi to end of element

147
xp(bi, bji) = LegD(I * 3 - 2) + LegD(I * 3 - 1) * (r3) * D1 + _

LegD(I * 3) * (r5) / 2 * (3 * D1 ^ 2 - 1)
Next bji
End Sub
C.18 Solve the Compliance Matrix
The subroutine SolveCompliance uses the finite element loading data and dis-
placement data to evaluate the full compliance matrix.
Private Sub SolveCompliance()

’now solve the Cmatrix
SMatCopy Xr, comp
mi = Lg3L * 3
For bi = 1 To mi
For bji = bi + 1 To mi
If comp(bi, bji) <> 0 Then _
comp(bji, bi) = 0.5 * (comp(bi, bji) + comp(bji, bi))
comp(bi, bji) = comp(bji, bi) ’average
Next
Next
mi = mi + Lg3L - 1 ’for 1st constraint set (continuity)

mi = mi + Lg3L - 1 ’for 2nd constraints (slope)
If mi > Lg3L * 3 Then

ReDim Preserve comp(1 To Lg3L * 3, 1 To mi)
’have to build comp transposed due to redim rules
For bi = 1 To (Lg3L - 1)
comp(bi * 3 - 2, 3 * Lg3L + bi) = 1
comp(bi * 3 - 1, 3 * Lg3L + bi) = r3
comp(bi * 3, 3 * Lg3L + bi) = r5
comp(bi * 3 + 1, 3 * Lg3L + bi) = -1
148
comp(bi * 3 + 2, 3 * Lg3L + bi) = r3

comp(bi * 3 + 3, 3 * Lg3L + bi) = -r5
Next bi
End If
If mi > 4 * Lg3L Then

For bi = 1 To (Lg3L - 1)
comp(bi * 3 - 2, 4 * Lg3L - 1 + bi) = 0#
comp(bi * 3 - 1, 4 * Lg3L - 1 + bi) = r3
comp(bi * 3, 4 * Lg3L - 1 + bi) = 3 * r5
comp(bi * 3 + 1, 4 * Lg3L - 1 + bi) = 0#
comp(bi * 3 + 2, 4 * Lg3L - 1 + bi) = -r3
comp(bi * 3 + 3, 4 * Lg3L - 1 + bi) = 3 * r5
Next bi
End If
ReDim LegD(1 To mi)

For bi = 1 To Lg3L
LegD(bi * 3 - 2) = u ’set the constant unit displacement
Next bi
SMatCopy XA, comp
MatT XA ’now xa is the normal comp matrix
D1 = QRSolve(XA, LegD, LegL)

lso = locS
GAULEG YBl, YBu, 250, tye, txe

ReDim YP(1 To 250)
mi = -1
For bi = 1 To 250
Do While tye(bi) > Yloc(mi + 2)
mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)

149
If I > ne Then I = ne: mi = mi - 2: Exit Do

Loop
d4 = Yloc(mi + 2) - tye(bi) ’distance from yi to end of element
YP(bi) = LegL(I * 3 - 2) + LegL(I * 3 - 1) * r3 * D1 + _

LegL(I * 3) * r5 / 2 * (3 * D1 ^ 2 - 1)
Next bi
frm3DGraph.Show
Plot frm3DGraph, tye, YP, "Location along face (m)", "Loading (N/m)", "p=KTC*u"
End Sub
C.19 Reset the Loading Data
The subroutine NotConverged resets the loading data for the next iteration for a
single loading and line of contact.
Private Sub NotConverged()

If loadflag = 0 Then loadflag = 2
’Ri = Wei * bj * le for these below

Wei(0, 1) = -1 / 6: Wei(0, 3) = -1 / 6: Wei(0, 6) = -1 / 6
Wei(0, 8) = -1 / 6: Wei(0, 4) = 1 / 3: Wei(0, 5) = 1 / 3
Wei(0, 2) = 1 / 2: Wei(0, 7) = 1 / 2 ’first leg loading
Wei(1, 1) = 0: Wei(1, 3) = 0: Wei(1, 6) = 0: Wei(1, 8) = 0

Wei(1, 4) = 0: Wei(1, 5) = 0
Wei(1, 2) = r3 / 6: Wei(1, 7) = -r3 / 6 ’ second loading only
Wei(2, 1) = r5 / 30: Wei(2, 3) = r5 / 30: Wei(2, 6) = r5 / 30

Wei(2, 8) = r5 / 30: Wei(2, 4) = -r5 / 15: Wei(2, 5) = -r5 / 15
Wei(2, 2) = 0: Wei(2, 7) = 0 ’third loading only
150
End Sub
C.20 Save the Critical Iteration Data
The subroutine FillKTCys saves the current iteration data for further use by the
program or in case of process interruption.
Private Sub FillKTCys()

LoadText = ""
AddLine LoadText, CStr(locS)
AddLine LoadText, CStr(locS2)
AddLine LoadText, CStr(sn)
AddLine LoadText, CStr(ne)
DprintMat LegL, , ln1, True

AddLine LoadText, ln1
DprintMat Yloc, , ln1, True

AddLine LoadText, ln1
AddLine LoadText, CStr(YBl)

AddLine LoadText, CStr(YBu)
AddLine LoadText, CStr(e)
AddLine LoadText, CStr(alphaOmega)
End Sub
C.21 Determine the Boundary of the Unmodified Region
The subroutine FindLocEnd determines the edge of the nominally unmodified

contact region.
Private Sub FindLocEnd()

151
Dim Xplot#(), Yplot#()

YBl = Dmax1((-delta * Beta / 2 - sn * Beta) / gamma, -delta * Beta / 2 / gamma)
YBu = Dmin1((delta * Beta / 2 - sn * Beta) / gamma, delta * Beta / 2 / gamma)
If locS = 0 Then locS = 1 + 2 * nem
If sn = 0 Then
e = 0#
ReDim YP(1 To 250)
ye = 0#
mi = -1
For bi = 1 To 250
mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)
If I = 0 Then I = 1
Loop
YP(bi) = LegL(I * 3 - 2) + LegL(I * 3 - 1) * (r3) * D1 + _
LegL(3 * I) * (r5) / 2 * (3 * D1 ^ 2 - 1)
ye = ye + YP(bi) * txe(bi)
Next bi
alphaOmega = 1 / Cos(Psib) * ye / u
alphaOmega0 = alphaOmega
Else
Dim dele#, ml#, mu#, aOL%, aOB#
ReDim YP(1 To 4)
ReDim tye(1 To 4)
e = YBu
tye(1) = YBl: tye(2) = YBu

If aOB = 0 Then aOB = alphaOmega0
ai = 1
152
at = 50
thu = 1
Do
mi = locS
I = 1
GAULEG YBl, YBu, 250, Yr, txe
dele = Yr(1) - YBl
Dim Wjs As Double
ReDim YP(0 To 251)

ye = 0#
For bi = 0 To 251
If bi < 1 Then
d8 = YBl
ElseIf bi > 250 Then
d8 = YBu
Else
d8 = Yr(bi)
End If
Do While d8 > Yloc(mi + 2) And mi < locS2 - 2

mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)
If I = 0 Then I = 1
If I > (locS2 - locS) / 2 Then I = (locS2 - locS) / 2
Loop
d4 = Yloc(mi + 2) - d8 ’distance from yi to end of element
LegL(3 * I) * (r5) / 2 * (3 * D1 ^ 2 - 1)
If bi > 0 And bi < 251 Then ye = ye + YP(bi) * txe(bi)
Next bi
Wjs = 1 / Cos(Psib) * ye / u
alphaOmega = Wjs
153
ml = (YP(1) - YP(0)) / dele / u

mu = (YP(250) - YP(251)) / dele / u
A = (-ml + mu) * 0.5

b = (YP(0) + YP(251))/u
c = aOB * (1 - Abs(sn) / delta) - alphaOmega
dele = (-b + Sqrt(Abs(b ^ 2 - 4 * A * c))) / (2 * A)
YBl = YBl + dele

YBu = YBu - dele
Loop Until Abs(dele / YBu) < 0.00000000000001
e = YBu - e
End If
End Sub
C.22 Determine the Edge Extension
The subroutine FindTheta determines the new boundary for a line of contact
under increased loading. At each load step a new extension is found.
Private Sub FindTheta()

Dim Ws#, auH#, Omega#, aOL#, mu#, dele#, ml#
ReDim KTC(1 To 2)
ReDim YP(1 To 250)
ye = 0#
I = 1
If bc(1) <> bc(2) Then
GAULEG bc(1), bc(2), 250, tye, txe
mi = locS
For bi = 1 To 250
mi = mi + 2
154
I = CInt((1 + mi - locS) / 2 + 0.1)

Loop
LegL(3 * I) * (r5) / 2 * (3 * D1 ^ 2 - 1)
Next bi
End If
alphaOmega = 1 / Cos(Psib) * ye / u
’ as calculated in the unmodified region for this s and u
ai = 10
bgn:
DoEvents
If InStr(1, ",9,14,19,24,", Chr(44) + CStr(loadflag) + Chr(44)) Then
ye = 0#
I = 1
mi = locS
For bi = 1 To 250
Do While tye(bi) > Yloc(mi + 2) And mi < locS2
mi = mi + 2
I = CInt((1 + mi - locS) / 2 + 0.1)
Loop
LegL(3 * I) * (r5) / 2 * (3 * D1 ^ 2 - 1)
155
Next bi
’ should be the integral over the full area -- (alpha Omega_fi) * loading
dW = 1 / Cos(Psib) * ye
tu = du
Else
dW = alphaOmega * u
tu = u
End If
Omega = alphaOmega / alpha

auH = alpha * u / h
If sn / delta <= auH / 2 And sn / delta >= -auH / 2 Then

Ws = Omega * h * (1 + betas * (auH) ^ (1 + ks)) * _
(auH - 0.25 * auH ^ 2 - (sn / delta) ^ 2)
ElseIf sn / delta <= -auH / 2 And sn / delta >= -1 + auH / 2 Then
Ws = Omega * h * auH * (1 + betas * (auH) ^ (1 + ks)) * _(1 + (sn / delta))
ElseIf sn / delta <= -1 + auH / 2 And sn / delta >= -1 - auH / 2 Then
Ws = Omega * h / 2 * (1 + betas * (auH) ^ (1 + ks)) * (1 + auH / 2 + _
(sn / delta)) ^ 2
ElseIf sn / delta >= auH / 2 And sn / delta <= 1 - auH / 2 Then
Ws = Omega * h * auH * (1 + betas * (auH) ^ (1 + ks)) * (1 + (sn / delta))
ElseIf sn / delta >= 1 - auH / 2 And sn / delta <= 1 + auH / 2 Then
Ws = Omega * h / 2 * (1 + betas * (auH) ^ (1 + ks)) * (1 + auH / 2 + _
(sn / delta)) ^ 2
End If
Debug.Print "Ws, alphaOmega * u , dWj"

Debug.Print Ws, alphaOmega * u, dW
iu = 0
I = 0
d12 = 1#

ml = (1 - 1 / (r2)) * thetai
156
YBl = bc(1) - ml
YBu = bc(2) + ml
Do
KTC(1) = SavedLoad(YBl, locS, Yloc, LegL) / u
KTC(2) = SavedLoad(YBu, locS, Yloc, LegL) / u
thetai = 2 * (Ws - (dW)) / (KTC(1) + KTC(2)) / (tu)
ml = (1 - 1 / (r2)) * thetai
dele = mu - thetai
YBl = bc(1) - ml * d12
YBu = bc(2) + ml * d12
I = I + 1
ye = mu
mu = thetai
iu = iu + 1
Loop While Abs(dele) > Abs(mu) * 0.00000001

Else
Dim thin$, thout$
If mu < 2 Then
reinit:
If yBll > yBul Then SwapValues yBll, yBul
YBl = yBll - thetai
YBu = yBul + thetai
Else
thin = thin + " , " + CStr(thetai)
End If
mu = 0
a0 = 0
A1 = 0
d8 = 0
d4 = dW
157
Dim ThetaAve#
Dim Thet#(1 To 100), iti%, therei%
Dim TaveL#
iti = 1
d20 = yBul - yBll
Do
ml = thetai
KTC(1) = SavedLoad(YBl, locS, Yloc, LegL) / u
KTC(2) = SavedLoad(YBu, locS, Yloc, LegL) / u
thetai = 2 * (Ws - (dW)) / (KTC(1) + KTC(2)) / (tu)
Thet(iti) = thetai
iti = iti + 1
If iti > 100 Then iti = 1
TaveL = ThetaAve
ThetaAve = Ave(Thet)
Dim C0#, C1#, C2#

mu = mu + 1
C0 = C1
C1 = C2
C2 = ml
a0 = A1
A1 = A2
A2 = thetai
If a0 <> 0 And CSng(C1) <> CSng(C0) Then

b1 = (A1 - a0) / (C1 - C0) ’line from two points (slope)
B2 = A1 - C1 * b1 ’y-int B2 = a0 - C0 * b1
If b1 <> 1 Then thetai = B2 / (1 - b1) ’self consistent point
158
End If
YBl = yBll - thetai

YBu = yBul + thetai
Loop While Abs(ml - thetai) > Abs(thetai) / 100000000#
thout = thout + " , " + CStr(thetai)
aOL = 0
eti = eti + 1
If Abs(dele - thetai) < thetai / 100000000# Then
GoTo fini
ElseIf Len(thin) > 35 Then
thetai = ConvTheta(thin, thout)
End If
If eti > 100 Then
’ assume that we are reaching the limits of our convergence ability
If ai = 2 Then
’take average of thetai
Str2ValMat txe, thout
thetai = Ave(txe)
GoTo fini
Else
ai = ai - 1
eti = 0
End If
End If
If Len(thin) > 500 Then thout = "": thin = ""
dele = thetai
b3 = thetai
YBl = yBll - thetai

YBu = yBul + thetai
159
GoTo bgn
End If
Debug.Print "YBu, YBl, thetai, alphaOmega"
Debug.Print YBu, YBl, thetai, alphaOmega
End Sub
160
C.23 Determine Step Size
The subroutine ConvTheta determines a best fit solution to the step function
between deflection cases.
Public Function ConvTheta(ThetaIn As String, ThetaOut As String)

Dim tI#(), TT#(), A#, b#, lI%
mid:
Call Str2ValMat(tI, ThetaIn, , , -ai)

A = UBound(tI)
Call Str2ValMat(TT, ThetaOut, , , -A)
A = Min0((A), UBound(TT))
lI = 0
Dim tit#(), ki%, mi%
ReDim tit(1 To A)
For ki = 1 To A
tit(ki) = TT(lI + ki)
Next ki
For ki = A To 1 Step -1
For mi = 1 To ki - 1
If tI(mi) > tI(ki) Then
b = tI(ki)
tI(ki) = tI(mi)
tI(mi) = b
b = tit(ki)
tit(ki) = tit(mi)
tit(mi) = b
End If
Next mi
Next ki
161
’ determine the number of similar digits

b = DigComp(tI(A), tI(1)) ’Log((tI(A) - tI(1)) / Abs(tI(CInt(A / 2)))) / Log(10)
b = CInt(10 - b) ’number of regression points to use.
If b < A Then
ai = Dmax1(2, ai - 1)
If ai < A Then eti = 0:
Debug.Print "Regression N trigger"
End If
ki = A
Dim Ri#
If A = 2 Then
Call LineTwoPoints(tI, tit, A, b)
Else
RegressionLin tI, tit, ki, b, A, Ri
End If
ConvTheta = b / (1 - A)
End Function
162
C.24 Apply Default Loading
The subroutine ZeroLoad creates an approximate initial load case for the compli-
ance iteration sequence.
Public Function ZeroLoad(ByVal y#) As Double

Dim A1#, b1#
If F = 0.02266557 Then
If y < 0.0115 Then ’
A1 = -3000
b1 = 35
ZeroLoad = A1 * y + b1 ’N/m
Else
ZeroLoad = 10#
End If
ElseIf F = 0.06096 Then
If y < 0.031 Then
A1 = -1130
b1 = 35
ZeroLoad = A1 * y + b1 ’N/m
Else
ZeroLoad = 10#
End If
Else
ZeroLoad = 10#
End If
End Function
163
C.25 Open A Saved Data Set
The subroutine SavedLoad opens a current loading save file for use by the pro-
gram.
Public Function SavedLoad(ByVal yi#, sl%, Yloc#(), LegL#()) As Double

Dim N#, m%, N2#, K%, yil%, yiu%, maxL#
If even(sl) Then
sl = sl - 1
End If
yil = UBound(Yloc)
K = UBound(LegL)
If CInt(yi) = yi And yi > 0 Then

N = Yloc(yi)
m = CInt((1 + yi - sl) / 2 - 0.1)
Else
N = yi
m = 1 - nem
yiu = 3
Do While N > Yloc(yiu) And yiu < yil
m = m + 1
yiu = m * 2 + sl
Loop
yi = m * 2 + sl - 1
End If
Dim rflag%
If m <= 0 Then m = 1: rflag = 1
If m > K / 3 Then m = K / 3: rflag = 1
yiu = m * 2 + sl
yil = yiu - 2
N = Yloc(yiu) - N ’distance from yi to end of element

N2 = Yloc(yiu) - Yloc(yil) ’total length of element
164
N = -N / N2 * 2 + 1 ’eta on the -1 to 1 range
SavedLoad = LegL(m * 3 - 2) + LegL(m * 3 - 1) * r3 * N + _

LegL(m * 3) * r5 / 2 * (3 * N ^ 2 - 1)
End Function
C.26 Rotate the Node Plane
The subroutine NodesRot takes an axial position of nodes and rotates the set to
introduce a helix to the model.
Public Sub NodesRot(Nodes() As Double, ang#)

Dim X#, y#, J%, ju%
ju = UBound(Nodes, 2) ’ nodes)
For J = 1 To ju
X = Nodes(1, J)
y = Nodes(2, J)
Nodes(1, J) = X * Cos(ang) - y * Sin(ang)
Nodes(2, J) = X * Sin(ang) + y * Cos(ang)
Next J
End Sub
165
References
[1] J. E. Shigley and J. Uicker, Theory of Machines and Mechanisms, McGraw-Hill

Inc., 1980.
[2] W. A. Tuplin, Involute Gear Geometry, Chatto and Windus, 1962.
[3] S. L. Harris, Proceedings of the Institute of Mechanical Engineers 172, 87 (1958).
[4] R. W. Gregory, S. L. Harris, and R. G. Munro, Proceedings of the Institute of

Mechanical Engineers (1963).
[5] W. D. Mark, Journal of the Acoustical Society of America 63, 1409 (1978).
[6] W. D. Mark, Journal of the Acoustical Society of America 66, 1758 (1979).
[7] T. G. Chondros, The development of machine design as a science from classical

times to modern era, in International Symposium on History of Machines and
Mechanisms: Proceedings of HMM 2008, edited by H.-S. Yan and M. Ceccarelli,
pages 59–68, New York, NY, 2009, Springer.
[8] T. Koetsier, Mechanism and machine science: Its history and its identity, in Inter-
national Symposium on History of Machines and Mechanisms - Proceedings HMM
2000, edited by M. Ceccarelli, pages 5–26, Norwell, MA, 2000, Kluwer Academic
Publishers.
[9] A. D. Dimarogonas, The Machine: A Historical Introduction, chapter 0, pages 1–33,

John Wiley and Sons, Inc., New York, NY, 2001.
[10] L. P. Euler, Supplementum de figura dentium rotarium, 1765.
[11] J. D. Smith, Gear Noise and Vibration, Marcel Dekker, Inc., New York, NY, second
edition, 2003.
[12] H. Walker, The Engineer 166, 409 (1938).
[13] D. B. Welbourn, Fundamental knowledge of gear noise - a survey, in Noise and

Vibrations of Engines and Transmissions, pages 9–14, London, 1979, Cranfield.
[14] D. R. Houser, Gear Noise Sources and Their Prediction Using Mathematical Models,
chapter 16, pages 213–222, Society of Automotive Engineers, 1990.
166
[15] W. D. Mark, Elements of Gear Noise Prediction, chapter 21, pages 735–769, John
Wiley and Sons, Inc., 1992.
[16] R. G. Munro, A review of the theory and measurement of gear transmission er-
ror, in Proceedings of the Institution of Mechanical Engineers: First International
Concerence: Gearbox Noise and Vibration, pages 3–10, Institution of Mechanical
Engineers, 1990.
[17] H. N. Özgüven and D. R. Houser, Journal of Sound Vibration 125, 71 (1988).
[18] R. G. Munro, N. Yildirim, and D. M. Hall, Optimum profile relief and transmission
error in spur gears, in Proceedings of the Institution of Mechanical Engineers: First
International Concerence: Gearbox Noise and Vibration, pages 35–42, Institution
of Mechanical Engineers, 1990.
[19] R. G. Munro, N. Yildirim, and D. L. Hall, Gearbox noise and vibration , 35 (1990).
[20] A. Kahraman and G. W. Blankenship, Journal of Mechanical Design 121, 313

(1999).
[21] F. L. Litvin et al., Computer methods in applied mechanics and engineering 192,
3619 (2003).
[22] F. L. Litvin, J. Lu, D. P. Townsend, and M. Howkins, Mechanism and machine

theory 34, 123 (1999).
[23] F. L. Litvin, A. Fuentes, A. Demenego, D. Vecchiato, and Q. Fan, New developments

in the design and generation of gear drives, in Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, volume
215, pages 747–757, Professional Engineering Publishing, 2001.
[24] S. Chang, D. R. Houser, and J. Harianto, Tooth flank corrections of wide face
width helical gears that account for shaft deflections, in Proceedings of the 9th
International Power Transmission and Gearing Conference (AGMA Fall Technical
Conference), Chicago, IL, ASME / AGMA.
[25] P. Wagaj and A. Kahraman, Impact of tooth profile modifications on the transmis-
sion error excitation of helical gear pairs, in Proceedings of ESDA2002: 6th Biennial
Conference on Engineering Systems Design and Analysis, 2002.
[26] A. Andersson and L. Vedmar, Journal of Sound and Vibration 260, 195 (2003).
167
[27] D. R. Houser, Research of the Ohio State University gear dynamics and gear noise
research laboratory, in Gearbox noise and vibration: First International Conference,
9-11 April 1990, Churchill College, University of Cambridge, page 19, Published for
the Institution of Mechanical Engineers by Mechanical Engineering Publications
Limited, 1990.
[28] D. R. Houser, J. Harianto, N. Iyer, J. Josephson, and B. Chandrasekaren, A

multi variable approach to determining the ‘best’ gear design, in Proceedings of
DETC’2000 ASME Power transmission and gearing conference, 2000.
[29] S. Sundaresan, K. Ishii, and D. R. Houser, Journal of Mechanical Design 113, 318
(1991).
[30] P. Velex and M. Maatar, Journal of Sound and Vibration 191, 629 (1996).
[31] A. Kahraman and G. W. Blankenship, Journal of Mechanical Design 121, 112

(1999).
[32] N. H. Özgüven and D. R. Houser, Journal of Sound and Vibration 121, 383 (1988).
[33] G. Liu and R. G. Parker, Journal of Mechanical Design 130, 121402 (2008).
[34] R. G. Parker, S. M. Vijayakar, and T. Imajo, Journal of Sound and Vibration 237,
435 (2000).
[35] R. V. Baud and R. E. Peterson, Mechanical Engineering 51, 653 (1929).
[38] A. Y. Attia, ASME Journal of Engineering for Industry 86, 333 (1964).
[39] C. Weber, The deformation of loaded gears and the effect on their load carry-
ing capacity, Technical Report 3, British Department of Scientific and Industrial
Research, 1949.
[40] R. W. Cornell, ASME Journal of Mechanical Design 103, 447 (1981).
[41] R. J. Roark and W. C. Young, Roark’s Formulas of Stress and Strain, McGraw-Hill
Inc., sixth edition, 1989.
[42] T. F. Coney and A. Seirig, Journal of Engineering for Industry , 1115 (1973).
168
[43] D. R. Houser et al., Static and dynamic transmission error measurements and predic-
tion for spur and helical gear sets, in Power Transmission and Gearing Conference,
ASME Design Engineering Division.
[44] M. Stegemiller and D. R. Houser, 3d analysis of the base flexibility of gear teeth,
in Proceedings of the 1989 Power Transmission and Gearing Conference, Chicago,
1989.
[45] D. B. Yakubek and D. R. Houser, 3D deflection analysis of gear teeth using both
FEA and a tapered plate approximation, Technical report, AGMA, 1985, AGMA
paper 85FTM4.
[46] W. B. Rockwood et al., Finite elements applied to the nonlinear calculation of tooth
loads and root stresses in high power density gearing, in AIAA/SAE/ASME/ASEE
23rd Joint Propulsion Conference, 1987, Report No. AAIA-87-1811.
[47] M. Umeyama et al., Effects of gear dimension and tooth surface modifications of
the loaded transmission error of a helical gear pair, in Power Transmission and
Gearing Conference, pages 725–732, ASME Design Engineering Division, 1996.
[48] J. J. Coy and C. Hu-Chih, ASME Journal of Mechanical Design 104, 759 (1982).
[49] W. E. Welker, A method for finite element computation of the contact deformation
of gear teeth, Master’s thesis, The Pennsylvania State University, 1996.
[50] E. M. Jankowich, A method for finite-element computation of the contact defor-

mation of helical gear teeth, Master’s thesis, The Pennsylvania State University,
1997.
[51] M. F. Alulis, A method for finite element computation of compliance influence

functions of helical gear teeth, Master’s thesis, The Pennsylvania State University,
1999.
[52] J. E. Shigley and C. R. Mischke, Mechanical Engineering Design, McGraw-Hill

Inc., fifth edition, 1989.
[53] W. D. Mark, Gear Noise, chapter 36, pages 36.1–36.22, McGraw-Hill Inc., 1991.
[54] M. L. Baxter, Jr., Basic theory of gear-tooth action and generation, in Gear
Handbook: The Design, Manufacture, and Application of Gears, edited by D. W.
Dudley, chapter 1, McGraw-Hill Inc., New York, first edition, 1962.
169
[55] D. W. Kammler, A First Course in Fourier Analysis, Prentice Hall, Inc., 2000.
[56] G. Liu and R. G. Parker, Journal of Sound and Vibration 320, 1039 (2009).
[57] J. D. Smith, Helical gear vibration excitation with misalignment, in Proceedings of

the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering
Science, volume 208, pages 71–79, Professional Engineering Publishing, 1994.
[58] J. D. Smith, Proceedings of the Institution of Mechanical Engineers, Part C: Journal

of Mechanical Engineering Science 212, 217 (1998).
[59] F. L. Litvin, Q. Lian, and A. L. Kapelevich, Computer Methods in Applied Me-

chanics and Engineering 188, 363 (2000).
[60] E. W. Weisstein, CRC Concise Encyclopedia of Mathematics, CRC Press, 2003.

Vita
Cameron Paul Reagor was born in Great Falls, Montana, on March 11, 1977. He
graduated from Shelby High School in Shelby, Montana, in May, 1995. He received a
Bachelor of Science degree, with honors, in Mechanical Engineering from the Montana
State University in May, 1999. He enrolled in the Ph.D. program in Acoustics at The
Pennsylvania State University in August, 1999.
He currently owns an acoustics consulting business in Montana.

Camp HD Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Camp HD Thesis

Uploaded by

Copyright:

Available Formats

The Pennsylvania State University

The Graduate School

AN OPTIMAL GEAR DESIGN METHOD FOR MINIMIZATION

c 2010 Cameron P. Reagor

Submitted in Partial Fulfillment

*Signatures are on file in the Graduate School.

Fluctuation in static transmission error is the accepted principal cause of vibration

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Chapter 2. Previous Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Chapter 4. The Main Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4.6.2 Continuing the Expansion . . . . . . . . . . . . . . . . . . . . 42

Chapter 5. The Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Chapter 6. Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 92

Appendix B. Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Appendix C. The Analysis Primary Code . . . . . . . . . . . . . . . . . . . . . . 104

C.12 Mesh the Finite Element Model . . . . . . . . . . . . . . . . . . . . . 123

5.1 Test Design Gear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

A.1 Summary of General Gearing Notation . . . . . . . . . . . . . . . . . . . 97

3.3 Dual Construction: for a set of local pressure angles, φi , locations on a

4.1 Frequency domain theoretical gear noise is bound to integer multiples of

5.1 A rendered example of the test gear. . . . . . . . . . . . . . . . . . . . 54

5.17 The loading curves for s = 0.0017967 m . . . . . . . . . . . . . . . . . . 63

2.1 History of Gearing

2.2 Noise in Gearing

As a practical matter, audible gear noise resulting from transmission error is

100 200 300 400 500 600 700

Fourier Transmission Error Spectrum

0 20 40 60 80 100 120 140

2.3 Current Gear Noise State of the Art

2.4 Development of a Complete Solution

2.5 Present Work

A necessary prerequisite for understanding gear transmission error analysis is un-

3.1 Gearing Constructs

3.2 Mathematical Preliminaries

3.2.1 The Involute Curve

x = Rb cos(²) + Rb ² sin(²) and y = Rb sin(²) − Rb ² cos(²). (3.1)

3.2.2 Angle Relations

tan(ψb ) = cos(φ) tan(ψ). (3.2)

The normal pressure angle can be found by

tan(φn ) = cos(ψ) tan(φ). (3.3)

and is derived in Appendix B.

3.2.3 Gear Coordinates

z can be defined in terms of s and y:

s = z/β − γy/β = Rb (² − ²0 ) ∓ tan(ψb )y. (3.16)

Fig. 3.10. Gear tooth ranges, L and D, relative to tooth rotation.

As is evident from Figure 3.10, D is defined through L:

L = L(1) + L(2) = [Rb (²a − ²p )](1) + [Rb (²a − ²p )](2) . (3.18)

3.2.4 Regular Gear Relations

and the circular pitch,

and dedendum radius,

3.3 Transmission Error

There are two components of transmission error [5]:

ζ(x) = uj (x, y) + ηj (x, y). (3.27)

ζ(x) = ζ (1) (x) + ζ (2) (x), (3.28)

l = y sec ψb ∴ dl = sec(ψb )dy. (3.32)

3.4 The Tooth Force Equation

The total force transmitted by tooth pair j,

kj (l, l0 ; x) is defined as the deformation at l on the line of contact on tooth pair j

The tooth force equation,

is developed in terms of the tooth stiffness and can be expressed as

In the familiar x and y terms the force equation is

Using Equation 3.43 and solving gives the stiffness formulation,